6 Principles to Make Self-Driving Cars Work for Cities
June 27, 2016 9:47 PM Subscribe
Asking cars to be fully automated (as opposed to having a human in the operating loop), will push things back by years, if not decades. And I'm rather dim on the prospect of even the mostly self-guided ones showing up soon.
While these suggestions seem like simple things from a public health perspective, they miss some of the practical/legal/economic incentives that are going to come into play with fully fledged self-driving cars.
For instance: as parking in downtown cores becomes more expensive and batteries become cheaper/better, what are the suburb dwellers going to do with their cars during the day? It could easily be that it's cheaper to have your uber-efficient electric car drive around all day, rather than pay parking. This would create maddening congestion levels and contribute to heavy power consumption.
Also, if people aren't required to do anything in their auto-pods, it could encourage more far-flung housing bases. This would also increase congestion on highway routes, even if each car was more efficiently driven than a human could manage.
posted by cult_url_bias at 10:46 PM on June 27, 2016
While these suggestions seem like simple things from a public health perspective, they miss some of the practical/legal/economic incentives that are going to come into play with fully fledged self-driving cars.
For instance: as parking in downtown cores becomes more expensive and batteries become cheaper/better, what are the suburb dwellers going to do with their cars during the day? It could easily be that it's cheaper to have your uber-efficient electric car drive around all day, rather than pay parking. This would create maddening congestion levels and contribute to heavy power consumption.
Also, if people aren't required to do anything in their auto-pods, it could encourage more far-flung housing bases. This would also increase congestion on highway routes, even if each car was more efficiently driven than a human could manage.
posted by cult_url_bias at 10:46 PM on June 27, 2016
First, I think NACTO's great. Second, I think it's important to note that autonomous vehicles are both potentially the solution to most of our transportation problems and potentially the most destructive force to cities in history - I saw a presentation where the US DOE researchers ballparked fossil fuel consumption with AVs to be cut to 25% or increased to 300% of existing or anywhere in between. So policy makers need to get out in front of the technology.
I mostly agree with their points, and would add a 7th.
1. Cars should be fully autonomous, not partly
This is totally, 100% correct. Partly autonomous vehicles are deathtraps. On research test tracks, where participants are told this is experimental equipment and it's an obvious test situation, they still instantly trust the technology. (There's plenty of Tesla autopilot near miss videos on YouTube if you look.) You go from a situation like texting while driving - where driving is the important activity - to driving while texting, where the driving response is delayed until you're done writing the text. Have you ever been writing an email and someone comes into the office and you need just a second to finish the thought you're writing? And maybe you're a little rude or you don't realize it's your boss's boss, but you shut them out, thinking just let me finish this? It's exactly like that, except the person coming into your office is your car, announcing you're hurtling towards a situation at 60 mph that is so complicated and unusual it has no idea what to do.
2. Maximum speeds on city streets should not exceed 25 miles per hour
This is correct and should be done already, but the political will is difficult. Very difficult.
3. Reconsider highway expansions
Correct, but the capacity argument is well overstated. The capacity gains you can get from reasonable automation are on the order of 10-15 mph on a highway and decreasing following distance to maybe 1/3rd. Go look at a highway sometime - everybody is already speeding by 10-15 mph and following too closely, so the bulk of the capacity improvement will be to reproduce existing use and capacity of the highways without killing 30,000 Americans a year.
4. Require vehicles to gather important public data
This is good, with the caveat that for a long while this will only tell us where rich techies want to go and not be representative of the general population - you already see this problem with cities using Strava data for cycling. This can convert the digital divide into physical infrastructure.
5. Study how vehicle automation can improve transit
This is the killer app in some senses - 65-75% of the cost of operating a transit system is in wages, largely driver's wages. While automating paratransit and first mile / last mile are both important, if you can slash transit operator wages you can massively increase service provided for the same cost. And the relationship between transit service and ridership is nonlinear; Seattle and Honolulu provide double the hours of transit service per capita of cities like Austin and Cleveland, but have 2.7x the ridership (and fare revenue). Austin and Cleveland provide double the hours of service of cities like Little Rock and Lancaster CA, with again roughly 2.8x the ridership.
6. People should use automated vehicles to build the cities they want, not build cities to accommodate automated vehicles
Great in theory, again this is a political will question.
The 7th I would add is maybe more specific:
7. Zero occupant travel should only be legal for fleet operations.
One of the best policy levers in North America for managing auto use is parking supply and costs, and if cars are allowed to drive themselves empty, then the roads will instantly be clogged with empty cars driving back out of downtown to park at home for free. In somewhere like Manhattan, it would be cheaper to drive to work and then tell your car to just circle the block for eight hours than to pay for parking. Except if even a fraction of people do this, then it's obviously gridlock.
Where autonomous vehicles could provide real benefits is in fleet operations like Uber - they won't replace all uses of cars (I think most parents like having diapers and Cheerios and Frozen DVDs and the like stashed in their cars, for example), but this is a chance to provide a step towards better multimodal mobility - a car there if it's needed but not if you can do without. So fleet operation (including transit, obvs.) without passengers makes sense, but if we let individuals drive their own cars empty, we're creating so much more traffic (and carbon emissions) than we already do, and what we already do is unsustainable and unlivable.
posted by Homeboy Trouble at 11:05 PM on June 27, 2016 [26 favorites]
I mostly agree with their points, and would add a 7th.
1. Cars should be fully autonomous, not partly
This is totally, 100% correct. Partly autonomous vehicles are deathtraps. On research test tracks, where participants are told this is experimental equipment and it's an obvious test situation, they still instantly trust the technology. (There's plenty of Tesla autopilot near miss videos on YouTube if you look.) You go from a situation like texting while driving - where driving is the important activity - to driving while texting, where the driving response is delayed until you're done writing the text. Have you ever been writing an email and someone comes into the office and you need just a second to finish the thought you're writing? And maybe you're a little rude or you don't realize it's your boss's boss, but you shut them out, thinking just let me finish this? It's exactly like that, except the person coming into your office is your car, announcing you're hurtling towards a situation at 60 mph that is so complicated and unusual it has no idea what to do.
2. Maximum speeds on city streets should not exceed 25 miles per hour
This is correct and should be done already, but the political will is difficult. Very difficult.
3. Reconsider highway expansions
Correct, but the capacity argument is well overstated. The capacity gains you can get from reasonable automation are on the order of 10-15 mph on a highway and decreasing following distance to maybe 1/3rd. Go look at a highway sometime - everybody is already speeding by 10-15 mph and following too closely, so the bulk of the capacity improvement will be to reproduce existing use and capacity of the highways without killing 30,000 Americans a year.
4. Require vehicles to gather important public data
This is good, with the caveat that for a long while this will only tell us where rich techies want to go and not be representative of the general population - you already see this problem with cities using Strava data for cycling. This can convert the digital divide into physical infrastructure.
5. Study how vehicle automation can improve transit
This is the killer app in some senses - 65-75% of the cost of operating a transit system is in wages, largely driver's wages. While automating paratransit and first mile / last mile are both important, if you can slash transit operator wages you can massively increase service provided for the same cost. And the relationship between transit service and ridership is nonlinear; Seattle and Honolulu provide double the hours of transit service per capita of cities like Austin and Cleveland, but have 2.7x the ridership (and fare revenue). Austin and Cleveland provide double the hours of service of cities like Little Rock and Lancaster CA, with again roughly 2.8x the ridership.
6. People should use automated vehicles to build the cities they want, not build cities to accommodate automated vehicles
Great in theory, again this is a political will question.
The 7th I would add is maybe more specific:
7. Zero occupant travel should only be legal for fleet operations.
One of the best policy levers in North America for managing auto use is parking supply and costs, and if cars are allowed to drive themselves empty, then the roads will instantly be clogged with empty cars driving back out of downtown to park at home for free. In somewhere like Manhattan, it would be cheaper to drive to work and then tell your car to just circle the block for eight hours than to pay for parking. Except if even a fraction of people do this, then it's obviously gridlock.
Where autonomous vehicles could provide real benefits is in fleet operations like Uber - they won't replace all uses of cars (I think most parents like having diapers and Cheerios and Frozen DVDs and the like stashed in their cars, for example), but this is a chance to provide a step towards better multimodal mobility - a car there if it's needed but not if you can do without. So fleet operation (including transit, obvs.) without passengers makes sense, but if we let individuals drive their own cars empty, we're creating so much more traffic (and carbon emissions) than we already do, and what we already do is unsustainable and unlivable.
posted by Homeboy Trouble at 11:05 PM on June 27, 2016 [26 favorites]
Cars are going to be totally driverless because that's the only system that will work safely and efficiently. The cars already exist. We just need to use them.
As for things like cars circling endlessly -- when we talk about driverless cars, we enter robot territory. These are robots running approved programs -- approved by you, but also approved by the government. They have to be running the robot equivalent of local traffic regulations. The approved programs for robot cars need to make sure our robots don't do stuff we (individually and collectively) don't want to allow them to do, and to make sure they work together, not against one another (and against us).
A city could, for example, allow a driverless private car to drop off its occupants and then automatically claim, pay for, and drive directly to the nearest free parking space until its owner calls for it. If the car didn't actually park because the owner recalled it before it parked, the owner would still have paid for the time between claiming the spot and recalling the car.
A city could also track all vehicle movements and charge tolls for vehicles exceeding normal approved use (for example, distance per day within the city), so an endlessly circling vehicle, whether occupied or not, would pay more than a vehicle that just parked and waited.
Or your private robot car could be programmed to give rides to others (maybe to total strangers, maybe from an approved pool of people) when you aren't using it, so it would earn a little of its keep as a robot taxi and reduce the demand for additional private cars on your streets.
The main dangers to society posed by driverless cars are longer commutes and urban sprawl -- people living a couple of hours from work because they can sleep or watch movies in their cars while they commute to and from work. But if we simultaneously improve city living (by allowing only quiet, clean, safe electric cars on the city streets) and use tolls and highway planning to shape long-distance traffic, city homes will remain in demand and the cars that drive into cities will at least be quiet, clean, and safe.
posted by pracowity at 1:34 AM on June 28, 2016 [4 favorites]
As for things like cars circling endlessly -- when we talk about driverless cars, we enter robot territory. These are robots running approved programs -- approved by you, but also approved by the government. They have to be running the robot equivalent of local traffic regulations. The approved programs for robot cars need to make sure our robots don't do stuff we (individually and collectively) don't want to allow them to do, and to make sure they work together, not against one another (and against us).
A city could, for example, allow a driverless private car to drop off its occupants and then automatically claim, pay for, and drive directly to the nearest free parking space until its owner calls for it. If the car didn't actually park because the owner recalled it before it parked, the owner would still have paid for the time between claiming the spot and recalling the car.
A city could also track all vehicle movements and charge tolls for vehicles exceeding normal approved use (for example, distance per day within the city), so an endlessly circling vehicle, whether occupied or not, would pay more than a vehicle that just parked and waited.
Or your private robot car could be programmed to give rides to others (maybe to total strangers, maybe from an approved pool of people) when you aren't using it, so it would earn a little of its keep as a robot taxi and reduce the demand for additional private cars on your streets.
The main dangers to society posed by driverless cars are longer commutes and urban sprawl -- people living a couple of hours from work because they can sleep or watch movies in their cars while they commute to and from work. But if we simultaneously improve city living (by allowing only quiet, clean, safe electric cars on the city streets) and use tolls and highway planning to shape long-distance traffic, city homes will remain in demand and the cars that drive into cities will at least be quiet, clean, and safe.
posted by pracowity at 1:34 AM on June 28, 2016 [4 favorites]
I've mentioned in other threads that my nightmare scenario, besides increased congestion or fuel consumption, is cars that follow you around shouting advertisements at you and trying to persuade you to get into them for a free ride to somewhere you might spend money. Maybe there need to more broadly be rules about under what conditions a car can suggest destinations or alternate routes.
Also, it seems as though in many places you'll have roads that can function with a smaller footprint / maintained surface... in fact, come to think of it, you could reduce the number of lanes in almost any road or street as a trade off with speed if the cars going in different directions can negotiate and maneuver around each other, right? Even if usage patterns stayed the same, in many cities I've been to when you're traveling inwards or outwards with commuter traffic the other side of the highway is empty, so with every vehicle computer-coordinated you could use half the number of lanes and just change the direction of lanes as the day goes on.
So governments everywhere may find themselves holding gold mines of prime real estate, and it seems like we need to get ahead of the curve on that issue before it all gets sold off for closing yearly budget gaps and / or embezzlement.
I would think in particular city dwellers who want their cities to be more walkable and bikeable should make sure there are rules in place that will prioritize any freed-up streets for those uses before they're liquidated.
On the radio last year I heard a story concerning rails-to-trails projects, where a court ruling decided that in some cases land that was seized by eminent domain and given to railroad companies had to be returned to the original owner once it permanently would no longer be used for a railway. So the parallel situation could create all kinds of fun. Maybe another vector for indigenous reclamation of land rights, when a road is old enough?
posted by XMLicious at 2:25 AM on June 28, 2016 [2 favorites]
Also, it seems as though in many places you'll have roads that can function with a smaller footprint / maintained surface... in fact, come to think of it, you could reduce the number of lanes in almost any road or street as a trade off with speed if the cars going in different directions can negotiate and maneuver around each other, right? Even if usage patterns stayed the same, in many cities I've been to when you're traveling inwards or outwards with commuter traffic the other side of the highway is empty, so with every vehicle computer-coordinated you could use half the number of lanes and just change the direction of lanes as the day goes on.
So governments everywhere may find themselves holding gold mines of prime real estate, and it seems like we need to get ahead of the curve on that issue before it all gets sold off for closing yearly budget gaps and / or embezzlement.
I would think in particular city dwellers who want their cities to be more walkable and bikeable should make sure there are rules in place that will prioritize any freed-up streets for those uses before they're liquidated.
On the radio last year I heard a story concerning rails-to-trails projects, where a court ruling decided that in some cases land that was seized by eminent domain and given to railroad companies had to be returned to the original owner once it permanently would no longer be used for a railway. So the parallel situation could create all kinds of fun. Maybe another vector for indigenous reclamation of land rights, when a road is old enough?
posted by XMLicious at 2:25 AM on June 28, 2016 [2 favorites]
Oh fachrissakes not this again.
Autonomous vehicles are an impractical solution that hardly anybody wants to a problem that doesn't exist. It's silly.
posted by GallonOfAlan at 3:27 AM on June 28, 2016 [3 favorites]
Autonomous vehicles are an impractical solution that hardly anybody wants to a problem that doesn't exist. It's silly.
posted by GallonOfAlan at 3:27 AM on June 28, 2016 [3 favorites]
And nowhere in the recommendations is there any suggestion on what to do with/for the millions of workers who are going to lose jobs in our new wonderful auto'mated society.
posted by Beholder at 3:37 AM on June 28, 2016 [2 favorites]
posted by Beholder at 3:37 AM on June 28, 2016 [2 favorites]
Transit may be the first huge win. Not specifically in wages saved but in load balancing. There are huge spikes during commute and other periods, some are managed over the years by the transit organization building a city wide load model but it certainly fails frequently due to localized spikes. It is just impractical to have a half dozen buses on standby can you imagine the budget discussions with the mayor if there were a 6-10 drivers that just sat around every day? But a few (perhaps smaller) buses, possibly staged near different trouble spots could be ready to go into service and smooth loads. Huge win. No more standing on jammed buses.
The other reason transit may be an early candidate is the routes are known and finite. A ultra detailed mapping survey of a cities bus routes would be more practical than "every road in the world".
posted by sammyo at 4:29 AM on June 28, 2016 [1 favorite]
The other reason transit may be an early candidate is the routes are known and finite. A ultra detailed mapping survey of a cities bus routes would be more practical than "every road in the world".
posted by sammyo at 4:29 AM on June 28, 2016 [1 favorite]
Oh fachrissakes not this again.
Gun deaths vs car.
posted by sammyo at 4:31 AM on June 28, 2016 [5 favorites]
Gun deaths vs car.
posted by sammyo at 4:31 AM on June 28, 2016 [5 favorites]
This is the killer app in some senses - 65-75% of the cost of operating a transit system is in wages, largely driver's wages. While automating paratransit and first mile / last mile are both important, if you can slash transit operator wages you can massively increase service provided for the same cost.
From a service and safety perspective, what I'd love is to have the buses become autonomous and the drivers turned into conductors. There are a lot of reasons to have a human staff person on board, but not much reason for them to be driving the vehicle if it can drive itself.
posted by Dip Flash at 5:18 AM on June 28, 2016 [8 favorites]
From a service and safety perspective, what I'd love is to have the buses become autonomous and the drivers turned into conductors. There are a lot of reasons to have a human staff person on board, but not much reason for them to be driving the vehicle if it can drive itself.
posted by Dip Flash at 5:18 AM on June 28, 2016 [8 favorites]
Every time I hear proponents talk about the improvements they hope to see out of automated cars--especially those who think that driverless cars will lead to greatly increased car-sharing--I get the impression that most of what they want would be better provided by really extensive personal rapid transit systems.
Does anyone know of any studies of the absolute minimum scale for a rail system? I'd nearly swear that you could move 4-6 person PRT cabs along a single cable-stayed grade-separated rail ("Monorail!") using street-level infrastructure that wasn't much more intrusive than existing light posts. Sure seems like it could be affordable to build even by asphalt and gravel standards, and confinement to a rail and grade-separation from all possible dangers that aren't under the same control (also from us endangered fragile fleshbag pedestrians) would simplify the problem of safe control algorithms almost to the point of being trivial.
posted by CHoldredge at 5:34 AM on June 28, 2016
Does anyone know of any studies of the absolute minimum scale for a rail system? I'd nearly swear that you could move 4-6 person PRT cabs along a single cable-stayed grade-separated rail ("Monorail!") using street-level infrastructure that wasn't much more intrusive than existing light posts. Sure seems like it could be affordable to build even by asphalt and gravel standards, and confinement to a rail and grade-separation from all possible dangers that aren't under the same control (also from us endangered fragile fleshbag pedestrians) would simplify the problem of safe control algorithms almost to the point of being trivial.
posted by CHoldredge at 5:34 AM on June 28, 2016
Autonomous cars are basically driverless taxis. It doesn't make sense to own one. When you are driven to work the car isn't going to go park itself, it will go pick up someone else. Sure some people will own them, but that will be for status, or if you live in a remote area.
posted by bhnyc at 5:54 AM on June 28, 2016 [8 favorites]
posted by bhnyc at 5:54 AM on June 28, 2016 [8 favorites]
Autonomous vehicles are an impractical solution that hardly anybody wants to a problem that doesn't exist. It's silly.
35,000 people dying per year in automobile crashes in the USA is hardly a problem that doesn't exist -and the most common cause is "driver inattention."
Yes, many of these deaths can be prevented with street design too.
But people ARE developing autonomous cars - so it just seems sensible to have some public policy that addresses them to ensure that they're a net positive regardless of whether or not one thinks that they are a practical solution.
posted by entropone at 6:10 AM on June 28, 2016 [7 favorites]
35,000 people dying per year in automobile crashes in the USA is hardly a problem that doesn't exist -and the most common cause is "driver inattention."
Yes, many of these deaths can be prevented with street design too.
But people ARE developing autonomous cars - so it just seems sensible to have some public policy that addresses them to ensure that they're a net positive regardless of whether or not one thinks that they are a practical solution.
posted by entropone at 6:10 AM on June 28, 2016 [7 favorites]
Asking cars to be fully automated (as opposed to having a human in the operating loop), will push things back by years, if not decades.
That's a feature, not a bug.
If we want partial automation, we need to consider how drivers will react when the system degrades to a lesser degree of automation, or requires human input. Mode confusion is a huge problem among highly-experienced airline pilots, and there's no reason to suggest that it wouldn't pose similar problems in cars.
posted by schmod at 6:12 AM on June 28, 2016 [4 favorites]
That's a feature, not a bug.
If we want partial automation, we need to consider how drivers will react when the system degrades to a lesser degree of automation, or requires human input. Mode confusion is a huge problem among highly-experienced airline pilots, and there's no reason to suggest that it wouldn't pose similar problems in cars.
posted by schmod at 6:12 AM on June 28, 2016 [4 favorites]
Oh fachrissakes not this again.
Gun deaths vs car.
So, the answer is clearly automated guns.
posted by leotrotsky at 6:29 AM on June 28, 2016 [3 favorites]
Gun deaths vs car.
So, the answer is clearly automated guns.
posted by leotrotsky at 6:29 AM on June 28, 2016 [3 favorites]
Autonomous cars are basically driverless taxis. It doesn't make sense to own one.
It will be cheaper to buy and maintain your own car than pay the premiums required to catch a personal car during rush hour, and the premiums required to drop you off someplace out of the way. It will be a great choice for off-peak driving; errands where you need to haul cargo or kids.
Autonomous shuttles that can programmatically adjust their route to accommodate riders are a thing, and Local Motors has a cool design for one. Mass-transit solutions like this are a better idea than a zillion autonomous cabs whizzing about during the commute.
posted by Slap*Happy at 6:34 AM on June 28, 2016 [2 favorites]
It will be cheaper to buy and maintain your own car than pay the premiums required to catch a personal car during rush hour, and the premiums required to drop you off someplace out of the way. It will be a great choice for off-peak driving; errands where you need to haul cargo or kids.
Autonomous shuttles that can programmatically adjust their route to accommodate riders are a thing, and Local Motors has a cool design for one. Mass-transit solutions like this are a better idea than a zillion autonomous cabs whizzing about during the commute.
posted by Slap*Happy at 6:34 AM on June 28, 2016 [2 favorites]
If we want partial automation, we need to consider how drivers will react when the system degrades to a lesser degree of automation, or requires human input. Mode confusion is a huge problem among highly-experienced airline pilots, and there's no reason to suggest that it wouldn't pose similar problems in cars.
That's why Ospreys are so dangerous, as well. You do different stuff in 'helicopter' mode than you do in 'plane' mode, and confusion of the two can easily kill you.
In auto-drive mode, you shouldn't even be able to see a steering wheel, let alone use it. In self-drive, there should be multiple indications (visual, auditory, etc) that you alone are the driver now.
posted by leotrotsky at 6:36 AM on June 28, 2016 [1 favorite]
That's why Ospreys are so dangerous, as well. You do different stuff in 'helicopter' mode than you do in 'plane' mode, and confusion of the two can easily kill you.
In auto-drive mode, you shouldn't even be able to see a steering wheel, let alone use it. In self-drive, there should be multiple indications (visual, auditory, etc) that you alone are the driver now.
posted by leotrotsky at 6:36 AM on June 28, 2016 [1 favorite]
Autonomous cars are basically driverless taxis. It doesn't make sense to own one.
It will be cheaper to buy and maintain your own car than pay the premiums required to catch a personal car during rush hour, and the premiums required to drop you off someplace out of the way. It will be a great choice for off-peak driving; errands where you need to haul cargo or kids.
The answer is probably somewhere in between, either a fractional ownership or maybe a yearly allotment of use-hours. One issue is that demand is certainly spiky, but I think that's a problem that can be solved in ways other than full ownership; a pseudo minibus that can transport multiple riders might be one solution.
posted by leotrotsky at 6:40 AM on June 28, 2016
It will be cheaper to buy and maintain your own car than pay the premiums required to catch a personal car during rush hour, and the premiums required to drop you off someplace out of the way. It will be a great choice for off-peak driving; errands where you need to haul cargo or kids.
The answer is probably somewhere in between, either a fractional ownership or maybe a yearly allotment of use-hours. One issue is that demand is certainly spiky, but I think that's a problem that can be solved in ways other than full ownership; a pseudo minibus that can transport multiple riders might be one solution.
posted by leotrotsky at 6:40 AM on June 28, 2016
Gun deaths vs car.
Woah, what happened between 2006 and 2010?
posted by R a c h e l at 6:42 AM on June 28, 2016
Woah, what happened between 2006 and 2010?
posted by R a c h e l at 6:42 AM on June 28, 2016
Woah, what happened between 2006 and 2010?
A few hypotheses:
- Miles traveled peaked in 06-07 and then began falling slightly - that could contribute
- Cities started building safer bike and ped infrastructure
- Population shift back toward urban centers means fewer cars and safer conditions
posted by entropone at 6:56 AM on June 28, 2016 [1 favorite]
A few hypotheses:
- Miles traveled peaked in 06-07 and then began falling slightly - that could contribute
- Cities started building safer bike and ped infrastructure
- Population shift back toward urban centers means fewer cars and safer conditions
posted by entropone at 6:56 AM on June 28, 2016 [1 favorite]
“One thing about which fish Americans know exactly nothing is water autocentric society, since they have no anti-environment which would enable them to perceive the element they live in."
posted by entropicamericana at 7:08 AM on June 28, 2016 [1 favorite]
posted by entropicamericana at 7:08 AM on June 28, 2016 [1 favorite]
Another function which makes no sense economically but is still something I want: driverless cars for those who can't drive -- I know at least two people with low vision (much less than 20/50, in our jurisdiction) who would feel safer in a driverless car than in a car where they can't see the driver (and thus can't identify them later in a police line-up).
posted by Mogur at 7:20 AM on June 28, 2016 [2 favorites]
posted by Mogur at 7:20 AM on June 28, 2016 [2 favorites]
You know why driverless cars are absolutely coming, regardless of anything or anyone else? Because millions of old people -- and we all hope to count ourselves among their growing numbers one day -- want to go places but tend to have poor vision, poor hearing, poor reflexes, aching joints, and various other reasons why they shouldn't be careening down busy streets at the helm of a two-ton land yacht.
Old people alone are way more than enough financial and political force to guarantee that 100-percent driverless cars will be widely adopted as soon as they are available. "Hey, Sparky! Take me to the hairdressers already! And turn up the heat!"
posted by pracowity at 7:27 AM on June 28, 2016 [10 favorites]
Old people alone are way more than enough financial and political force to guarantee that 100-percent driverless cars will be widely adopted as soon as they are available. "Hey, Sparky! Take me to the hairdressers already! And turn up the heat!"
posted by pracowity at 7:27 AM on June 28, 2016 [10 favorites]
I'm disappointed my own preferred solution to the transportation problem has never gotten any traction: a high-speed mag-rail based system with personal cars you can own (but that could potentially be a lot cheaper without the need for internal combustion engines) and decouple from the track when you get to your particular stop for putting into storage (the car would also have a little battery driven electric motor for driving slowly to and from rail access points and for parking, so we could all still have somewhere to hang our invidious bumper stickers).
posted by saulgoodman at 7:36 AM on June 28, 2016 [1 favorite]
posted by saulgoodman at 7:36 AM on June 28, 2016 [1 favorite]
Autonomous cars are basically driverless taxis. It doesn't make sense to own one.
People speculate about this but If I want to get to work at peak usage time it probably makes sense to own one. Most cars during rush hour will be used only during rush hour, so the marginal cost will be high. Basically in owning my own car I'm paying for the capital cost of the car and gas, same as anyone I pay to move me around. I don't need to pay for the "extra distance" it travels at the start of the commute to pick me up or risk there not being a car available or able to get to me. The win for me is mostly I can read or work or watch a video if I'm stuck in traffic.
If I have a job where parking is bad, I might win by taking a driverless car--but of course that increases traffic in the most congested areas, as cars now both come and go and both ends of the commute.
Does anyone know of any studies of the absolute minimum scale for a rail system? I'd nearly swear that you could move 4-6 person PRT cabs along a single cable-stayed grade-separated rail ("Monorail!") using street-level infrastructure that wasn't much more intrusive than existing light posts.
Minimum scale for what? To be self-supporting? Every study I've seen summarized (I don't read primary literature) says it generally requires larger scale and more density than in most American cities, let alone suburbs. Which is why nearly everything is subsidized, including buses that use existing roads. In San Mateo county, where I live, it seems around 80% of the revenue is from sources other than passenger fares.
posted by mark k at 7:48 AM on June 28, 2016
People speculate about this but If I want to get to work at peak usage time it probably makes sense to own one. Most cars during rush hour will be used only during rush hour, so the marginal cost will be high. Basically in owning my own car I'm paying for the capital cost of the car and gas, same as anyone I pay to move me around. I don't need to pay for the "extra distance" it travels at the start of the commute to pick me up or risk there not being a car available or able to get to me. The win for me is mostly I can read or work or watch a video if I'm stuck in traffic.
If I have a job where parking is bad, I might win by taking a driverless car--but of course that increases traffic in the most congested areas, as cars now both come and go and both ends of the commute.
Does anyone know of any studies of the absolute minimum scale for a rail system? I'd nearly swear that you could move 4-6 person PRT cabs along a single cable-stayed grade-separated rail ("Monorail!") using street-level infrastructure that wasn't much more intrusive than existing light posts.
Minimum scale for what? To be self-supporting? Every study I've seen summarized (I don't read primary literature) says it generally requires larger scale and more density than in most American cities, let alone suburbs. Which is why nearly everything is subsidized, including buses that use existing roads. In San Mateo county, where I live, it seems around 80% of the revenue is from sources other than passenger fares.
posted by mark k at 7:48 AM on June 28, 2016
…PRT…After reading Latour’s Aramis (self-link to my notes), I reflexively reject any mention of PRT after he showed it to be essentially impossible: replacing the flexibility of the human body during rush hour with the rigidity of hardshell pods will never work in a city, for reasons of basic physics. Any kind of train or bus handles this problem, while any kind of car suffers from it. Driverless cars will be a worse experience than buses. Driverless transit might be interesting.
posted by migurski at 8:00 AM on June 28, 2016 [2 favorites]
People speculate about this but If I want to get to work at peak usage time it probably makes sense to own one. Most cars during rush hour will be used only during rush hour, so the marginal cost will be high.
I have thought this myself, but wouldn't it also be easier to get people to carpool/use public transit if the whole system was coordinated based on real-time demand and specific locations like Uber or Google Maps, instead of following predetermined routes and schedules?
posted by designbot at 8:09 AM on June 28, 2016
I have thought this myself, but wouldn't it also be easier to get people to carpool/use public transit if the whole system was coordinated based on real-time demand and specific locations like Uber or Google Maps, instead of following predetermined routes and schedules?
posted by designbot at 8:09 AM on June 28, 2016
Woah, what happened between 2006 and 2010?
Cash for Clunkers - lots of vehicles without modern safety features were taken off the road and replaced with newer, safer cars.
posted by Slap*Happy at 8:13 AM on June 28, 2016 [1 favorite]
Cash for Clunkers - lots of vehicles without modern safety features were taken off the road and replaced with newer, safer cars.
posted by Slap*Happy at 8:13 AM on June 28, 2016 [1 favorite]
You know why driverless cars are absolutely coming, regardless of anything or anyone else? Because millions of old people
... will be warehoused waiting to die, unless they have a lot of money, just like today?
posted by benzenedream at 8:22 AM on June 28, 2016 [1 favorite]
... will be warehoused waiting to die, unless they have a lot of money, just like today?
posted by benzenedream at 8:22 AM on June 28, 2016 [1 favorite]
35,000 people dying per year in automobile crashes in the USA is hardly a problem that doesn't exist -and the most common cause is "driver inattention."
It's certainly possible that this sentence will predict the future, only swapping out "system failure" for the cause....
posted by GenjiandProust at 8:23 AM on June 28, 2016
It's certainly possible that this sentence will predict the future, only swapping out "system failure" for the cause....
posted by GenjiandProust at 8:23 AM on June 28, 2016
Who is the driver of an automated car? The manufacturer. Make sure (through legislation) that manufacturers are liable for crashes, and "system failure" will be rare.
posted by yesster at 8:28 AM on June 28, 2016
posted by yesster at 8:28 AM on June 28, 2016
Woah, what happened between 2006 and 2010?
This NHTSA report only looks at the decline in 2008, but attributes it to less discretionary driving as a result of economic recession. This was particularly marked among young people and in rural areas.
Cities started building safer bike and ped infrastructure
One interesting thing I did find was that find is that, while there has been a decline in bike/ped motor vehicle deaths (thanks, infrastructure and awareness campaigns!), those "nonoccupant" deaths have increased as a proportion of total motor vehicle deaths. If you look at this NHTSA report, the proportion of "Pedestrians, Bicyclists and Other Nonoccupants" as fatalities increased from 13% in 2004 to 17% in 2013. Motorcycle fatalities as a proportion of the whole also increased (9 to 14%), while the biggest decline was among passenger car fatalities (45 to 37%).
posted by Panjandrum at 8:30 AM on June 28, 2016 [3 favorites]
This NHTSA report only looks at the decline in 2008, but attributes it to less discretionary driving as a result of economic recession. This was particularly marked among young people and in rural areas.
The Bureau of Labor Statistics has reported on the younger workforce in the United States having significantly higher rates of unemployment as compared to older age groups. It seems reasonable to suggest that this has probably had a big impact in the travel, both discretionary and non-discretionary, among young drivers... In conclusion, the significant decline in fatalities in 2008 was driven by large decreases in crashes involving young drivers, multiple-vehicle crashes, and crashes occurring during weekends. Areas that experienced greater increases in unemployment rates also recorded higher decreases in fatalities. When areas are redefined to include buffer zones, fatalities in rural areas declined more significantly than the fatalities in the urban and suburban areas.On the other hand, this study compared injury and hospitalization between 2001-2005 and 2006-2010 in Maryland, and found a few reasons for an overall decline in injury/hospitalization rates. One area they found to be significant was that cars themselves have gotten safer.
Reductions in total crashes during the past 10 years probably reflect a multitude of targeted safety efforts, including roadway improvements, campaigns against drunk driving, and vehicle safety improvements, among others. Decreases in police-reported injury severity, and overall hospital admission rates, on the other hand, are most likely attributable to improvements in crashworthiness that have been introduced during the past 10–15 years.So it seems like the decline in deaths could be attributed to less driving, particularly by young people, coupled with a range of safety improvements.
Cities started building safer bike and ped infrastructure
One interesting thing I did find was that find is that, while there has been a decline in bike/ped motor vehicle deaths (thanks, infrastructure and awareness campaigns!), those "nonoccupant" deaths have increased as a proportion of total motor vehicle deaths. If you look at this NHTSA report, the proportion of "Pedestrians, Bicyclists and Other Nonoccupants" as fatalities increased from 13% in 2004 to 17% in 2013. Motorcycle fatalities as a proportion of the whole also increased (9 to 14%), while the biggest decline was among passenger car fatalities (45 to 37%).
posted by Panjandrum at 8:30 AM on June 28, 2016 [3 favorites]
Autonomous cars are basically driverless taxis. It doesn't make sense to own one.
I'm not so sure. One big reason people like owning a personal car all of their own is that it's just that - 100% at their disposal anytime they want it, with no waiting, no need for pre-planning and the ability to change their mind right up to the last minute. For autonomous cars to escape those same environment-damaging preferences would require a fundamental change in human nature.
Also, if it's 100% your car, you can keep its interior exactly as clean as you want. In the taxis we have now, the human driver can periodically clean up any mess his less considerate passengers may leave behind.
But what happens with a driverless taxi if someone pukes up in the back seat then jumps out without cleaning it up? Does the car have some sort of system which detects this and sends it back to fleet HQ for a clean-up? Add a "bad behaviour" fine to that passenger's credit card? Or does it it just continue to drive round all day with the puke slowly congealing in the back seat?
I've picked an extreme example to make my point, but I am genuinely curious about this. Travis Bickle's not the only human cab driver who's regularly had to cope with the bodily fluids some passengers leave behind, so how do supporters of driverless taxis envisage them being kept habitable throughout the long working day?
posted by Paul Slade at 8:54 AM on June 28, 2016 [4 favorites]
I'm not so sure. One big reason people like owning a personal car all of their own is that it's just that - 100% at their disposal anytime they want it, with no waiting, no need for pre-planning and the ability to change their mind right up to the last minute. For autonomous cars to escape those same environment-damaging preferences would require a fundamental change in human nature.
Also, if it's 100% your car, you can keep its interior exactly as clean as you want. In the taxis we have now, the human driver can periodically clean up any mess his less considerate passengers may leave behind.
But what happens with a driverless taxi if someone pukes up in the back seat then jumps out without cleaning it up? Does the car have some sort of system which detects this and sends it back to fleet HQ for a clean-up? Add a "bad behaviour" fine to that passenger's credit card? Or does it it just continue to drive round all day with the puke slowly congealing in the back seat?
I've picked an extreme example to make my point, but I am genuinely curious about this. Travis Bickle's not the only human cab driver who's regularly had to cope with the bodily fluids some passengers leave behind, so how do supporters of driverless taxis envisage them being kept habitable throughout the long working day?
posted by Paul Slade at 8:54 AM on June 28, 2016 [4 favorites]
Maybe there will be a switch like in gas station bathrooms - "Flip this switch if service is required."
posted by entropone at 9:01 AM on June 28, 2016
posted by entropone at 9:01 AM on June 28, 2016
Make sure (through legislation) that manufacturers are liable for crashes, and "system failure" will be rare.
Ah, so if the technological fantasy is not sufficient, just apply more market policy fantasy.
posted by nom de poop at 9:07 AM on June 28, 2016 [2 favorites]
Ah, so if the technological fantasy is not sufficient, just apply more market policy fantasy.
posted by nom de poop at 9:07 AM on June 28, 2016 [2 favorites]
I'm disappointed my own preferred solution to the transportation problem has never gotten any traction: a high-speed mag-rail based system with personal cars you can own (but that could potentially be a lot cheaper without the need for internal combustion engines) and decouple from the track when you get to your particular stop for putting into storage
But that's exactly what a cross between Uber and autonomous cars would be. There is zero need for an actual rail infrastructure if the road can work as rigidly as the rails do in traffic control, which with autonomous but also swarm control can do. Making independent autonomous cars only does half the problem. Making it so that cars either cede control to the traffic computer/algorithm for that area or all talk to each other in some way is how it'd be great.
My own slice of that utopia would be that on highways, cars with similar destinations would hook or otherwise connect together into trains to save energy. So all the cars going more than 10 exits down would connect and use massively less energy to get to that destination, uncoupling a mile from their exit. With control of the car and the others around it you can get a LOT more cars into smaller spaces and transitions (where all the crap happens now) become no issue.
Unfortunately, to get the full gains, its an all or nothing solution. Manually driven cars will not be able to mix with autonomous. But I think that's an inevitable end point.
posted by Brockles at 9:12 AM on June 28, 2016 [2 favorites]
But that's exactly what a cross between Uber and autonomous cars would be. There is zero need for an actual rail infrastructure if the road can work as rigidly as the rails do in traffic control, which with autonomous but also swarm control can do. Making independent autonomous cars only does half the problem. Making it so that cars either cede control to the traffic computer/algorithm for that area or all talk to each other in some way is how it'd be great.
My own slice of that utopia would be that on highways, cars with similar destinations would hook or otherwise connect together into trains to save energy. So all the cars going more than 10 exits down would connect and use massively less energy to get to that destination, uncoupling a mile from their exit. With control of the car and the others around it you can get a LOT more cars into smaller spaces and transitions (where all the crap happens now) become no issue.
Unfortunately, to get the full gains, its an all or nothing solution. Manually driven cars will not be able to mix with autonomous. But I think that's an inevitable end point.
posted by Brockles at 9:12 AM on June 28, 2016 [2 favorites]
Make sure (through legislation) that manufacturers are liable for crashes, and "system failure" will be rare.
Of the cars will suddenly be 'much too difficult' to make, more likely.
posted by Brockles at 9:13 AM on June 28, 2016 [1 favorite]
Of the cars will suddenly be 'much too difficult' to make, more likely.
posted by Brockles at 9:13 AM on June 28, 2016 [1 favorite]
It's certainly possible that this sentence will predict the future, only swapping out "system failure" for the cause....
If one person texts while driving and crashes, they might decide never to text while driving again. But everyone else who texts while driving will carry on.
If an autonomous car crashes because of some system failure, that failure can be addressed and then every car using that system updated, in theory preventing that failure ever happening again.
posted by EndsOfInvention at 9:24 AM on June 28, 2016 [3 favorites]
If one person texts while driving and crashes, they might decide never to text while driving again. But everyone else who texts while driving will carry on.
If an autonomous car crashes because of some system failure, that failure can be addressed and then every car using that system updated, in theory preventing that failure ever happening again.
posted by EndsOfInvention at 9:24 AM on June 28, 2016 [3 favorites]
Speaking of utopias, whatever happened to us all working from home? Or not working as many hours? I'm ok with autonomous cars when I do need to drive, but solutions that don't require me to put on office clothes and commute at all would actually be better. This discussion assumes commuting from a dwelling to an office 5x a week is going to be the norm forever.
In fact, if I worked from home more, I would gladly give up the hassle and expense of owning my own car (or at least go from two to one) and order up a self-driving car when I did need to go somewhere. Provided that system was reliable.
posted by emjaybee at 9:28 AM on June 28, 2016 [4 favorites]
In fact, if I worked from home more, I would gladly give up the hassle and expense of owning my own car (or at least go from two to one) and order up a self-driving car when I did need to go somewhere. Provided that system was reliable.
posted by emjaybee at 9:28 AM on June 28, 2016 [4 favorites]
I realized the other day that self driving cars aren't a thing that will just show up all at once. It's a transition that's been in process since the automatic transmission was introduced. Cruise control, traction control, electronic power steering and throttles, and the like are all small towards full automation.
The first auto-pilot systems just held the stick straight for a few moments, now commercial aircraft basically fly themselves. The pilot is there in case something goes wrong and they monitor a bunch of stuff but if all goes according to plan, the pilot doesn't really do much. He's there so that if something goes wrong that the computer can't handle on it's own he can step in and get the aircraft safely to the ground. That's pretty much the only major thing we still need pilots for, otherwise commercial aircraft would get to 100% automation a lot sooner.
Cars just need to get to the shoulder, are generally easier to operate, and the stakes are typically lower so it's even easier to take the human out of the equation. We're not really waiting until cars are fully autonomous, we're waiting until they're autonomous enough that we need to start to change the way we drive. That, like 100% autonomous cars themselves, isn't something that gets put in place overnight but something that develops slowly over time in step with advances in automation.
posted by VTX at 10:08 AM on June 28, 2016 [2 favorites]
The first auto-pilot systems just held the stick straight for a few moments, now commercial aircraft basically fly themselves. The pilot is there in case something goes wrong and they monitor a bunch of stuff but if all goes according to plan, the pilot doesn't really do much. He's there so that if something goes wrong that the computer can't handle on it's own he can step in and get the aircraft safely to the ground. That's pretty much the only major thing we still need pilots for, otherwise commercial aircraft would get to 100% automation a lot sooner.
Cars just need to get to the shoulder, are generally easier to operate, and the stakes are typically lower so it's even easier to take the human out of the equation. We're not really waiting until cars are fully autonomous, we're waiting until they're autonomous enough that we need to start to change the way we drive. That, like 100% autonomous cars themselves, isn't something that gets put in place overnight but something that develops slowly over time in step with advances in automation.
posted by VTX at 10:08 AM on June 28, 2016 [2 favorites]
Autonomous vehicles are certainly coming in the near future, but as has been said, it's a very American and autocentric view that they are much help for crowded urban centers. Private industry will invest in drone cars, but governments need to invest in metros and trams.
posted by Steakfrites at 10:32 AM on June 28, 2016 [1 favorite]
posted by Steakfrites at 10:32 AM on June 28, 2016 [1 favorite]
... will be warehoused waiting to die, unless they have a lot of money, just like today?
Younger people will be happy to buy their elderly parents and grandparents cars that can take them all the places they currently have their children and grandchildren drive them (doctors, pharmacies, grocery shopping, visiting, etc.).
Also, there's economical carpooling. Two or three or four old people could go out together in one car and just enjoy the ride together.
And I do hope cities eventually evolve into the sort of place you could use smaller, cheaper city cars. Little economical electric two-seaters not made for highways but absolutely perfect for going a few blocks here and there.
posted by pracowity at 10:37 AM on June 28, 2016 [2 favorites]
Younger people will be happy to buy their elderly parents and grandparents cars that can take them all the places they currently have their children and grandchildren drive them (doctors, pharmacies, grocery shopping, visiting, etc.).
Also, there's economical carpooling. Two or three or four old people could go out together in one car and just enjoy the ride together.
And I do hope cities eventually evolve into the sort of place you could use smaller, cheaper city cars. Little economical electric two-seaters not made for highways but absolutely perfect for going a few blocks here and there.
posted by pracowity at 10:37 AM on June 28, 2016 [2 favorites]
There is zero need for an actual rail infrastructure if the road can work as rigidly as the rails do in traffic control,
You could design a much more energy efficient (and clean air) system using rails, though. That always seemed to me like it should be the main problem we're trying to solve. We could already without any new technology solve most traffic safety problems by all driving 25 mph or less at all times. Our expectations for how fast things should happen keeping getting reset to make us less and less patient. The faster we drive, the faster the pace of life, the more rushed we feel, the faster we drive... Etc. I don't know. Whatever we do, I just hope we don't make more and worse problems in the process as tends to happen from time to time.
posted by saulgoodman at 10:58 AM on June 28, 2016
You could design a much more energy efficient (and clean air) system using rails, though. That always seemed to me like it should be the main problem we're trying to solve. We could already without any new technology solve most traffic safety problems by all driving 25 mph or less at all times. Our expectations for how fast things should happen keeping getting reset to make us less and less patient. The faster we drive, the faster the pace of life, the more rushed we feel, the faster we drive... Etc. I don't know. Whatever we do, I just hope we don't make more and worse problems in the process as tends to happen from time to time.
posted by saulgoodman at 10:58 AM on June 28, 2016
This will be a public policy issue for generations. Glad to see people seriously working on it now. Cities are going to have to figure out how to attract people to live in them. A lot of people live in cities only because that's where the jobs are. Real estate's going to take a huge hit and, relatedly, so will the tax base.
posted by resurrexit at 11:06 AM on June 28, 2016
posted by resurrexit at 11:06 AM on June 28, 2016
You could design a much more energy efficient (and clean air) system using rails, though.
Electric cars and charging while publicly parked does exactly the same thing. Maybe inductive charging while on highways (so electrify the existing HOV lanes).
posted by Brockles at 11:10 AM on June 28, 2016
Electric cars and charging while publicly parked does exactly the same thing. Maybe inductive charging while on highways (so electrify the existing HOV lanes).
posted by Brockles at 11:10 AM on June 28, 2016
I thought the 25mph was a compromise.
It's surprisingly challenging to get mph adjusted downward, and at 25mph there's still more than a 10% chance you're going to kill someone if you hit them.
That's still too high for my tastes.
posted by aniola at 11:16 AM on June 28, 2016 [1 favorite]
It's surprisingly challenging to get mph adjusted downward, and at 25mph there's still more than a 10% chance you're going to kill someone if you hit them.
That's still too high for my tastes.
posted by aniola at 11:16 AM on June 28, 2016 [1 favorite]
Many people seem to be under the erroneous impression that the only negative externality of automobiles is the carbon emissions.
posted by entropicamericana at 11:18 AM on June 28, 2016 [3 favorites]
posted by entropicamericana at 11:18 AM on June 28, 2016 [3 favorites]
That's still too high for my tastes.
With modern cars (especially if we made them lighter, which they should be) that were following at a safe distance AND had a driver that is paying attention (ie either a good driver or a computer) 25mph is not very fast at all. You can stop a modern car extremely quickly from that speed.
The biggest issue in slow speed accidents is usually people not paying attention. In my experience, the slower people go, the more they look around, so going slower is not necessarily any kind of answer.
posted by Brockles at 11:25 AM on June 28, 2016 [1 favorite]
With modern cars (especially if we made them lighter, which they should be) that were following at a safe distance AND had a driver that is paying attention (ie either a good driver or a computer) 25mph is not very fast at all. You can stop a modern car extremely quickly from that speed.
The biggest issue in slow speed accidents is usually people not paying attention. In my experience, the slower people go, the more they look around, so going slower is not necessarily any kind of answer.
posted by Brockles at 11:25 AM on June 28, 2016 [1 favorite]
Speaking of utopias, whatever happened to us all working from home?
Marissa Myer did away with it to reduce headcount at Yahoo, and passed it off as a productivity-enhancer. Since she was a Big Deal at Google, a lot of companies copied her lead.
Of course, she's been something of a failure in revitalizing Yahoo, and the companies who kept work-at-home arrangements were filling vacancies and reducing overhead with no real dip in productivity. Pendulum is swinging back that way, and yes, it will be a blessing for the environment and human safety when commuting is avoided as much as possible by industry as a whole.
posted by Slap*Happy at 11:43 AM on June 28, 2016 [3 favorites]
Marissa Myer did away with it to reduce headcount at Yahoo, and passed it off as a productivity-enhancer. Since she was a Big Deal at Google, a lot of companies copied her lead.
Of course, she's been something of a failure in revitalizing Yahoo, and the companies who kept work-at-home arrangements were filling vacancies and reducing overhead with no real dip in productivity. Pendulum is swinging back that way, and yes, it will be a blessing for the environment and human safety when commuting is avoided as much as possible by industry as a whole.
posted by Slap*Happy at 11:43 AM on June 28, 2016 [3 favorites]
And I do hope cities eventually evolve into the sort of place you could use smaller, cheaper city cars
I've often thought it would be great if cars were more or less banned within city limits and people used a combination of transit and electric 'golf cart' type vehicles on a bikeshare-type model to get around.
Somewhat orthoganally, I was thinking the other day that it would be cool if Google leveraged their driving AI to create monitoring systems that could be used to detect traffic violations and ticket offenders, like amped-up red light cameras. There is currently so little enforcement of traffic laws, and fines are so low, that drivers basically just consider it 'bad luck' to be caught, and think nothing of engaging in careless, reckless or selfish behaviours that put people in danger or simply contribute to congestion. Automation seems to be the only possible way to scale enforcement to the volume of traffic on our roads in a cost-effective way.
posted by sevenyearlurk at 12:10 PM on June 28, 2016 [2 favorites]
I've often thought it would be great if cars were more or less banned within city limits and people used a combination of transit and electric 'golf cart' type vehicles on a bikeshare-type model to get around.
Somewhat orthoganally, I was thinking the other day that it would be cool if Google leveraged their driving AI to create monitoring systems that could be used to detect traffic violations and ticket offenders, like amped-up red light cameras. There is currently so little enforcement of traffic laws, and fines are so low, that drivers basically just consider it 'bad luck' to be caught, and think nothing of engaging in careless, reckless or selfish behaviours that put people in danger or simply contribute to congestion. Automation seems to be the only possible way to scale enforcement to the volume of traffic on our roads in a cost-effective way.
posted by sevenyearlurk at 12:10 PM on June 28, 2016 [2 favorites]
The correlation coefficient between the speed limit and the 85th percentile speed is .18 (IIRC) AKA almost none. Changing speed limits does VERY little to change the speeds people actually drive on a given road and speed isn't what's dangerous it's differences in speed that cause problems.
100% automated cars will up the correlation to 1 or near enough but until then, changing speed limits isn't going to be a good solution. You can't really control how hard the pedestrian gets hit but you can make it less likely that they'll be hit.
Cars are also going to get to 100% automation on the interstate long long before they're 100% automated 100% of the places they drive so you'll still be working around human drivers with pedestrians for a good while yet. Some semi-automated safety systems will help a lot with this too.
Speaking of utopias, whatever happened to us all working from home?
I'm home based and nearly all of the ~50+ people in my department work from home at least some of the time (and I think most work from home most of the time). I work for a large commercial bank that is a fortune 500 company.
posted by VTX at 12:13 PM on June 28, 2016
100% automated cars will up the correlation to 1 or near enough but until then, changing speed limits isn't going to be a good solution. You can't really control how hard the pedestrian gets hit but you can make it less likely that they'll be hit.
Cars are also going to get to 100% automation on the interstate long long before they're 100% automated 100% of the places they drive so you'll still be working around human drivers with pedestrians for a good while yet. Some semi-automated safety systems will help a lot with this too.
Speaking of utopias, whatever happened to us all working from home?
I'm home based and nearly all of the ~50+ people in my department work from home at least some of the time (and I think most work from home most of the time). I work for a large commercial bank that is a fortune 500 company.
posted by VTX at 12:13 PM on June 28, 2016
Somewhat orthoganally, I was thinking the other day that it would be cool if Google leveraged their driving AI to create monitoring systems that could be used to detect traffic violations and ticket offenders, like amped-up red light cameras. There is currently so little enforcement of traffic laws, and fines are so low, that drivers basically just consider it 'bad luck' to be caught, and think nothing of engaging in careless, reckless or selfish behaviours that put people in danger or simply contribute to congestion. Automation seems to be the only possible way to scale enforcement to the volume of traffic on our roads in a cost-effective way.
I don't even know where to start, but I'll try:
1. If your car is automated, what enforcement do you need? There are no decisions being made (by you)
2. Getting ticketed and fined is a lot more common when you're not white. Just saying.
3. Unless you are riding around with the cops all day how do you know lots of people are getting away with infractions?
4. Red light cameras are one of the most unpopular ideas out there, amping that up is not going to be greeted with anything but rage.
posted by emjaybee at 12:19 PM on June 28, 2016 [1 favorite]
I don't even know where to start, but I'll try:
1. If your car is automated, what enforcement do you need? There are no decisions being made (by you)
2. Getting ticketed and fined is a lot more common when you're not white. Just saying.
3. Unless you are riding around with the cops all day how do you know lots of people are getting away with infractions?
4. Red light cameras are one of the most unpopular ideas out there, amping that up is not going to be greeted with anything but rage.
posted by emjaybee at 12:19 PM on June 28, 2016 [1 favorite]
4. Red light cameras are one of the most unpopular ideas out there, amping that up is not going to be greeted with anything but rage.
I like the idea of finely focussed EMP weapons instead of cameras. Anyone that goes through a red light more than one second after it goes red gets their car zapped and essentially destroyed.
posted by Brockles at 12:32 PM on June 28, 2016 [2 favorites]
I like the idea of finely focussed EMP weapons instead of cameras. Anyone that goes through a red light more than one second after it goes red gets their car zapped and essentially destroyed.
posted by Brockles at 12:32 PM on June 28, 2016 [2 favorites]
How is the vehicle programmed to behave in a crash? Ethical problems of self-driving cars.
posted by lathrop at 12:35 PM on June 28, 2016
posted by lathrop at 12:35 PM on June 28, 2016
1. It will be a long, long time before all cars are automated.
2. Granted, although racist application of human enforcement would seem to be another argument in favour of automation.
3. I have eyes. Stand at any intersection in downtown Toronto for an hour and count just the number of people running yellows at the last microsecond, vs the number of police cars you see and the number of times someone gets stopped.
4. Constant rolling over on needed policy decisions to avoid 'enraging' drivers is what got us where we are today.
posted by sevenyearlurk at 12:36 PM on June 28, 2016 [2 favorites]
2. Granted, although racist application of human enforcement would seem to be another argument in favour of automation.
3. I have eyes. Stand at any intersection in downtown Toronto for an hour and count just the number of people running yellows at the last microsecond, vs the number of police cars you see and the number of times someone gets stopped.
4. Constant rolling over on needed policy decisions to avoid 'enraging' drivers is what got us where we are today.
posted by sevenyearlurk at 12:36 PM on June 28, 2016 [2 favorites]
3. Unless you are riding around with the cops all day how do you know lots of people are getting away with infractions?
I am a pedestrian, and I have eyes and a reasonably functional memory?
posted by GenjiandProust at 12:37 PM on June 28, 2016 [5 favorites]
I am a pedestrian, and I have eyes and a reasonably functional memory?
posted by GenjiandProust at 12:37 PM on June 28, 2016 [5 favorites]
The correlation coefficient between the speed limit and the 85th percentile speed is .18 (IIRC) AKA almost none. Changing speed limits does VERY little to change the speeds people actually drive on a given road
Sadly, true, which is why people suggest building streets to encourage/force drivers to drive more slowly.
and speed isn't what's dangerous it's differences in speed that cause problems.
wat
Is this some sort of "clever" STEM joke like it "It's not the fall it kills you, it's the sudden deceleration?"
posted by entropicamericana at 12:46 PM on June 28, 2016 [1 favorite]
Sadly, true, which is why people suggest building streets to encourage/force drivers to drive more slowly.
and speed isn't what's dangerous it's differences in speed that cause problems.
wat
Is this some sort of "clever" STEM joke like it "It's not the fall it kills you, it's the sudden deceleration?"
posted by entropicamericana at 12:46 PM on June 28, 2016 [1 favorite]
You need to back up a step. The first mistake you're making is to assume that traffic tickets are there to help enforce public safety, they're not, at least that's not how they're used. Cops write tickets to fund their department because a portion of those fines goes straight into that precinct's budget. So if you write enough tickets, your boss can afford to hire more cops.
Speed limits are the easiest one to pick on. The speed limit is typically set at the low end of the 85th percentile speed (the speed that 85% of drivers will drive on that stretch of road) and if you look at the fine schedule, you'll note that the fines take a BIG step up in cost at the TOP of the 85th percentile speed. If you dig around you can sometime find the traffic engineer reports from when the road was first built and you usually see some info about the expected 85th percentile speed there.
They figure out the speed that 85% of drivers will drive on that road and then set the speed so that you'll get a ticket if you drive that speed. The best part is that driving 5mph below the 85th percentile speed is just as dangerous and driving 5mph above. If speeding tickets were really about public safety, they'd set the speed limit near the top of 85th percentile speed range, increase the fines a LOT, and disconnect police funding from the fines.
You also want to keep in mind that the goal of the rules of the road isn't to follow the rules, it's to keep people safe. So if you're violating those rules without harming anyone, why should they be punished?
posted by VTX at 12:50 PM on June 28, 2016 [2 favorites]
Speed limits are the easiest one to pick on. The speed limit is typically set at the low end of the 85th percentile speed (the speed that 85% of drivers will drive on that stretch of road) and if you look at the fine schedule, you'll note that the fines take a BIG step up in cost at the TOP of the 85th percentile speed. If you dig around you can sometime find the traffic engineer reports from when the road was first built and you usually see some info about the expected 85th percentile speed there.
They figure out the speed that 85% of drivers will drive on that road and then set the speed so that you'll get a ticket if you drive that speed. The best part is that driving 5mph below the 85th percentile speed is just as dangerous and driving 5mph above. If speeding tickets were really about public safety, they'd set the speed limit near the top of 85th percentile speed range, increase the fines a LOT, and disconnect police funding from the fines.
You also want to keep in mind that the goal of the rules of the road isn't to follow the rules, it's to keep people safe. So if you're violating those rules without harming anyone, why should they be punished?
posted by VTX at 12:50 PM on June 28, 2016 [2 favorites]
Manually driven cars will not be able to mix with autonomous.
Manually driven cars currently mix with autonomous. They've been doing it for some years now. Autonomous cars are zipping along the normal public roads with normal cars right now.
And they may already be better drivers than most of the people sharing those roads.
posted by pracowity at 1:01 PM on June 28, 2016 [1 favorite]
Manually driven cars currently mix with autonomous. They've been doing it for some years now. Autonomous cars are zipping along the normal public roads with normal cars right now.
And they may already be better drivers than most of the people sharing those roads.
posted by pracowity at 1:01 PM on June 28, 2016 [1 favorite]
Speed kills, make no mistake about it. [more info + citations]
You also want to keep in mind that the goal of the rules of the road isn't to follow the rules, it's to keep people safe. So if you're violating those rules without harming anyone, why should they be punished?
What. This. I. Just.... *deep breath*
Let us extend this line of thinking to other areas like, oh, let's say nuclear power plants, workplace safety, and firearms and see where it takes us.
posted by entropicamericana at 1:28 PM on June 28, 2016 [2 favorites]
You also want to keep in mind that the goal of the rules of the road isn't to follow the rules, it's to keep people safe. So if you're violating those rules without harming anyone, why should they be punished?
What. This. I. Just.... *deep breath*
Let us extend this line of thinking to other areas like, oh, let's say nuclear power plants, workplace safety, and firearms and see where it takes us.
posted by entropicamericana at 1:28 PM on June 28, 2016 [2 favorites]
sevenyearlurk: "I was thinking the other day that it would be cool if Google leveraged their driving AI to create monitoring systems that could be used to detect traffic violations and ticket offenders, like amped-up red light cameras."
You follow anyone anyone with a regular commute for an hour and you will be able to write a ticket. Not only because humans aren't 100% accurate driving machines. But also because lots of traffic laws are discretionary. EG: it is common place to be ticketed for driving too fast for conditions even when driving well under the speed limit. Also to slow for conditions. Here they can cite you for stunting if you distract a pedestrian walking along side the road; usually used by cops who you have either pissed off in some legal manner or want to disperse youths.
posted by Mitheral at 2:25 PM on June 28, 2016
You follow anyone anyone with a regular commute for an hour and you will be able to write a ticket. Not only because humans aren't 100% accurate driving machines. But also because lots of traffic laws are discretionary. EG: it is common place to be ticketed for driving too fast for conditions even when driving well under the speed limit. Also to slow for conditions. Here they can cite you for stunting if you distract a pedestrian walking along side the road; usually used by cops who you have either pissed off in some legal manner or want to disperse youths.
posted by Mitheral at 2:25 PM on June 28, 2016
They've been doing it for some years now. Autonomous cars are zipping along the normal public roads with normal cars right now.
That's a misleading statement. It's more accurate to say 'A microscopically tiny (by proportion) number of manually monitored autonomous vehicles have been mixing with other traffic on extremely limited areas under test conditions'. It's hardly both existing alongside each other in equal proportion or with equal levels of control.
Speed kills, make no mistake about it.
People ALWAYS conflate this issue - speed and/or accidents. Speed kills, of course. A faster impact produces a more energetic collision which is statistically less survivable. But to extrapolate that to 'Speed also causes the accidents' is not valid. Given that an accident has been caused, speed is a major factor. But 'loss of control through excessive speed' is far more rarely the primary cause of accidents than repeating 'Speed Kills' blindly suggests.
posted by Brockles at 2:28 PM on June 28, 2016 [1 favorite]
That's a misleading statement. It's more accurate to say 'A microscopically tiny (by proportion) number of manually monitored autonomous vehicles have been mixing with other traffic on extremely limited areas under test conditions'. It's hardly both existing alongside each other in equal proportion or with equal levels of control.
Speed kills, make no mistake about it.
People ALWAYS conflate this issue - speed and/or accidents. Speed kills, of course. A faster impact produces a more energetic collision which is statistically less survivable. But to extrapolate that to 'Speed also causes the accidents' is not valid. Given that an accident has been caused, speed is a major factor. But 'loss of control through excessive speed' is far more rarely the primary cause of accidents than repeating 'Speed Kills' blindly suggests.
posted by Brockles at 2:28 PM on June 28, 2016 [1 favorite]
Road deaths? How about we drive less, educate drivers properly and police them properly maybe?
And as @Brockles correctly pointed out, driverless cars have in no way, shape or form been operating without human override on anything approaching real world roads.
And again the lazy bandying about of the term 'AI' that has been so prevalent recently. There IS NO AI. WE ARE AS FAR FROM AI AS EVER WE WERE.
posted by GallonOfAlan at 2:55 PM on June 28, 2016 [1 favorite]
And as @Brockles correctly pointed out, driverless cars have in no way, shape or form been operating without human override on anything approaching real world roads.
And again the lazy bandying about of the term 'AI' that has been so prevalent recently. There IS NO AI. WE ARE AS FAR FROM AI AS EVER WE WERE.
posted by GallonOfAlan at 2:55 PM on June 28, 2016 [1 favorite]
To expand on what Brockles said. Practically every accident in my jurisdiction gets speed listed as a contributing factor. If the cops get called to an accident scene they will site someone for driving too fast for conditions in all but the most freakish of accidents where there isn't some other bugaboo reason (distracted driving or impaired driving usually).
entropicamericana: "Let us extend this line of thinking to other areas like, oh, let's say nuclear power plants, workplace safety, and firearms and see where it takes us."
A lot of rules ostensibly put in place to enhance safety don't actually enhance safety. EG: I used to a travel a freeway at night where it wouldn't be unusual to encounter zero traffic over 100kms. Anyone you did encounter you saw kilometres away because of vehicle lights. A freeway designed with a travel speed of 140 km/hr. The speed limit at the time was 110 km/h. Was I endangering anyone travelling 120 km/h?
Does it change your answer if I note that the speed limit on that same highway was changed to 120 km/h last two years ago after a province wide review of highway speed limits?
Workplace safety example: I qualified for a driver's licence nearly 30 years ago and haven't had to show a proficiency to drive ever since; I never have to pass an examination as a journeyman electrician ever again. Yet I'm required to re-qualify on how to wear a fall arrest harness every two years. Do I somehow become a huge menace to safety on the job site at 24 months +1 day if I strap on a harness to work at heights? ~40% of workplace deaths in my province other than automobile accidents are the result of falls from less than 10'; the vast majority of them off of ladders. But we don't make workers have a ladder licence.
A lot of regulation in this area is designed to make the testing and certification agencies money without having a measurable effect on safety. There is a big push that is likely to succeed to extend re-certification requirements to three or four years in some of these areas. If successful what does that say about the tighter restrictions currently in place?
I'd be surprised if there was any area where this doesn't occur. Some like abortion are rife with "safety" laws that can be widely violated without any effect on safety. Hundreds of laws exist or have existed in the automobile sphere for "safety" that are/were all about regulatory capture; often 100% contradictory with laws in other jurisdictions.
posted by Mitheral at 3:06 PM on June 28, 2016
entropicamericana: "Let us extend this line of thinking to other areas like, oh, let's say nuclear power plants, workplace safety, and firearms and see where it takes us."
A lot of rules ostensibly put in place to enhance safety don't actually enhance safety. EG: I used to a travel a freeway at night where it wouldn't be unusual to encounter zero traffic over 100kms. Anyone you did encounter you saw kilometres away because of vehicle lights. A freeway designed with a travel speed of 140 km/hr. The speed limit at the time was 110 km/h. Was I endangering anyone travelling 120 km/h?
Does it change your answer if I note that the speed limit on that same highway was changed to 120 km/h last two years ago after a province wide review of highway speed limits?
Workplace safety example: I qualified for a driver's licence nearly 30 years ago and haven't had to show a proficiency to drive ever since; I never have to pass an examination as a journeyman electrician ever again. Yet I'm required to re-qualify on how to wear a fall arrest harness every two years. Do I somehow become a huge menace to safety on the job site at 24 months +1 day if I strap on a harness to work at heights? ~40% of workplace deaths in my province other than automobile accidents are the result of falls from less than 10'; the vast majority of them off of ladders. But we don't make workers have a ladder licence.
A lot of regulation in this area is designed to make the testing and certification agencies money without having a measurable effect on safety. There is a big push that is likely to succeed to extend re-certification requirements to three or four years in some of these areas. If successful what does that say about the tighter restrictions currently in place?
I'd be surprised if there was any area where this doesn't occur. Some like abortion are rife with "safety" laws that can be widely violated without any effect on safety. Hundreds of laws exist or have existed in the automobile sphere for "safety" that are/were all about regulatory capture; often 100% contradictory with laws in other jurisdictions.
posted by Mitheral at 3:06 PM on June 28, 2016
People ALWAYS conflate this issue - speed and/or accidents. Speed kills, of course. A faster impact produces a more energetic collision which is statistically less survivable. But to extrapolate that to 'Speed also causes the accidents' is not valid. Given that an accident has been caused, speed is a major factor. But 'loss of control through excessive speed' is far more rarely the primary cause of accidents than repeating 'Speed Kills' blindly suggests.
oh my god i smell burnt toast and my left arm just went numb
posted by entropicamericana at 3:11 PM on June 28, 2016 [1 favorite]
oh my god i smell burnt toast and my left arm just went numb
posted by entropicamericana at 3:11 PM on June 28, 2016 [1 favorite]
What. This. I. Just.... *deep breath*
So you're saying that if I'm on the interstate, all by myself, with no other cars in sight and I change lanes without a signal, that's just as deserving of punishment as if I did it in rush hour traffic?
posted by VTX at 3:12 PM on June 28, 2016 [1 favorite]
So you're saying that if I'm on the interstate, all by myself, with no other cars in sight and I change lanes without a signal, that's just as deserving of punishment as if I did it in rush hour traffic?
posted by VTX at 3:12 PM on June 28, 2016 [1 favorite]
And as @Brockles correctly pointed out, driverless cars have in no way, shape or form been operating without human override on anything approaching real world roads.
Half a year ago there was already at least one bus route in Triskala, Greece served by self-driving vehicles that mix with normal road traffic and don't have any attendants on board. The company's web site appears to indicate that the same pilot program was being run in four or five other cities around Europe.
The strenuous denialism that still shows up in these threads is bizarre given the copious numbers of autonomous vehicles operated by companies all over the world and the years and years worth of video of these vehicles in operation.
I mean sure, it could all be the result of a universal conspiracy, but short of explicitly proposing an explanation such as that, it's like someone who saw footage of the Trinity test of atomic weapons in 1945 saying "Well certainly it looks impressive but this was just a test, there's no evidence that a weapon like this would really be of use on the battlefield or that this technology will make a material difference in the way wars are fought."
posted by XMLicious at 3:24 PM on June 28, 2016 [2 favorites]
> So you're saying that if I'm on the interstate, all by myself, with no other cars in sight and I change lanes without a signal, that's just as deserving of punishment as if I did it in rush hour traffic?
Frankly, I'd punish that harder, but cars only deigning to signal when they see other cars around is a personal bugbear. Signals are especially important for the people you're not aware of, why have any threshold at all below which you decide not to use them? "No other cars in sight" shares a not insignificant area on the Venn diagram of road situations with "human being in your blind spot".
posted by lucidium at 3:28 PM on June 28, 2016 [4 favorites]
Frankly, I'd punish that harder, but cars only deigning to signal when they see other cars around is a personal bugbear. Signals are especially important for the people you're not aware of, why have any threshold at all below which you decide not to use them? "No other cars in sight" shares a not insignificant area on the Venn diagram of road situations with "human being in your blind spot".
posted by lucidium at 3:28 PM on June 28, 2016 [4 favorites]
The strenuous denialism that still shows up in these threads is bizarre given the copious numbers of autonomous vehicles operated by companies all over the world and the years and years worth of video of these vehicles in operation.
I think you need to look up what 'copious' means. You looked for examples and came up with one bus route in Greece, of all places and hints at merely 5 more? In all of global traffic? Pus maybe the 10-20 cars Google has? I'm probably being generous if I say there are 3-400 autonomous vehicles on public roads globally. Which is, as a proportion, almost negligible. Especially when most of them are still manned test beds.
This is not 'strenuous denial' it is realism. They are nowhere near as prevalent, nor as autonomous, as people were suggesting in this thread.
oh my god i smell burnt toast and my left arm just went numb
Are you here to have a discussion, or just drop weird sentences that are presumably meant to be the textual version of an exaggerated gasp into the thread?
posted by Brockles at 3:52 PM on June 28, 2016 [1 favorite]
I think you need to look up what 'copious' means. You looked for examples and came up with one bus route in Greece, of all places and hints at merely 5 more? In all of global traffic? Pus maybe the 10-20 cars Google has? I'm probably being generous if I say there are 3-400 autonomous vehicles on public roads globally. Which is, as a proportion, almost negligible. Especially when most of them are still manned test beds.
This is not 'strenuous denial' it is realism. They are nowhere near as prevalent, nor as autonomous, as people were suggesting in this thread.
oh my god i smell burnt toast and my left arm just went numb
Are you here to have a discussion, or just drop weird sentences that are presumably meant to be the textual version of an exaggerated gasp into the thread?
posted by Brockles at 3:52 PM on June 28, 2016 [1 favorite]
I think you need to look up what 'copious' means. You looked for examples and came up with one bus route in Greece, of all places and hints at merely 5 more?
No, off the top of my head without doing any research I could recall from a while ago an example easily disproving the claim by GallonOfAlan which I quoted and explicitly responded to that
driverless cars have in no way, shape or form been operating without human override on anything approaching real world roads"Copious" referred to the total number of companies around the globe that have developed systems or built upon those of other companies, and proved that this is a real and viable technology.
Yeah, this is like the year 1900 in the development of wireless telegraphy—it will be decades before devices based on the technology are in every household, we may not be seeing it in its final form, and whatever the equivalent is of Western Union may still be providing services based on the preceding century's technology a hundred-plus years in the future.
But people who are sticking their fingers in their ears and shouting "lalala it's all just a fiction!" and are still pushing back against and kicking and dragging their heels about anyone even considering the implications for human civilization of this development, are behaving absurdly.
We're well past the point where anyone can handwavily claim without evidence, as has been the fashion, that all tests by all companies and all researchers of all systems have been irredeemably flawed or contrived, because the technology has already actually been deployed.
posted by XMLicious at 5:15 PM on June 28, 2016 [3 favorites]
The strenuous denialism that still shows up in these threads ......people who are sticking their fingers in their ears and shouting "lalala it's all just a fiction!" ......We're well past the point where anyone can handwavily claim without evidence, as has been the fashion, that all tests by all companies and all researchers of all systems have been irredeemably flawed or contrived
See, now I am confused. Nobody here that I can see, is doing any of those things. You seem to be arguing against a position (with some badly placed/incorrect strong word usage, perhaps) that I haven't seen represented here.
posted by Brockles at 6:31 PM on June 28, 2016
See, now I am confused. Nobody here that I can see, is doing any of those things. You seem to be arguing against a position (with some badly placed/incorrect strong word usage, perhaps) that I haven't seen represented here.
posted by Brockles at 6:31 PM on June 28, 2016
Brockles - re-read the thread more closely. There are a lot of luddite denialists who think self-driving vehicles are "always ten years away" like fusion reactors. No, there are self-driving vehicles on common roadways operating in ordinary human-driven traffic, now, and they work and work well. It's one of the few examples of where tech is being dramatically under-promised and completely and thoroughly tested before being released to the general public. Everyone is following Google's lead, here, and their mandate stems back to the "Don't Be Evil" Google of yore, who never bothered to shed that cautious and thoughtful approach to the problem.
Yeah, machine learning is a thing, and it's enabling multiple players in the space to develop systems that work more than adequately without stepping on each others' toes. There are a lot of driverless cars, trucks and busses on public roads using a wide variety of systems and software, and their safe driving record, per mile, vastly exceeds humans. Vastly. Google alone has a few million miles worth of death-free driving under their belt at this time.
posted by Slap*Happy at 7:11 PM on June 28, 2016 [2 favorites]
Yeah, machine learning is a thing, and it's enabling multiple players in the space to develop systems that work more than adequately without stepping on each others' toes. There are a lot of driverless cars, trucks and busses on public roads using a wide variety of systems and software, and their safe driving record, per mile, vastly exceeds humans. Vastly. Google alone has a few million miles worth of death-free driving under their belt at this time.
posted by Slap*Happy at 7:11 PM on June 28, 2016 [2 favorites]
You may not have followed previous threads closely, Brockles. In that context, the hedging about whether driverless cars have ever operated in the "real world" and semantic quibbling about whether the term "AI" can or can't be bestowed on an automated system that drives a car in traffic (it does not matter at all what you call it) are part of a pattern.
So, I think it's worthwhile to clearly establish the full range of things that must be explained away and declared to not be "real" to assert that autonomous vehicle technology is vaporware or will never have any impact on society.
posted by XMLicious at 7:21 PM on June 28, 2016 [1 favorite]
So, I think it's worthwhile to clearly establish the full range of things that must be explained away and declared to not be "real" to assert that autonomous vehicle technology is vaporware or will never have any impact on society.
posted by XMLicious at 7:21 PM on June 28, 2016 [1 favorite]
The strenuous denialism that still shows up in these threads is bizarre given the copious numbers of autonomous vehicles operated by companies all over the world and the years and years worth of video of these vehicles in operation.
You linked to video of a self-driving bus that moves at "up to" 12 mph--that is, a vehicle that can be outpaced by a long-distance runner--and people who remain skeptical about the tech are engaging in "strenuous denialism"?
I started optimistic on self-driving tech and have gotten more and more skeptical the more I've learned. I think it will arrive sometime and I eagerly await the day. But my early optimism (because I personally hate driving and definitely trust engineering solutions more than my fellow human drivers) is tempered because all the tests rely on some asterisk or another. Google cars, for example, require careful mapping of the territory (according to wiki inch-scale resolution) and slow down drastically when something unexpected comes up. I've driven past them and yeah, they do drive that way too. Evidence of safety problems caused by partial autonomy also muddy the actual path to a big rollout.
Basically my dream was to get a commute where I no longer have to drive. I no longer think that's happening much before I retire (say, a decade or so) and there hasn't been anything in this thread to change my mind. Not sure what you're seeing that makes you think it's right around the corner as opposed to niche thing.
posted by mark k at 8:42 PM on June 28, 2016 [2 favorites]
You linked to video of a self-driving bus that moves at "up to" 12 mph--that is, a vehicle that can be outpaced by a long-distance runner--and people who remain skeptical about the tech are engaging in "strenuous denialism"?
I started optimistic on self-driving tech and have gotten more and more skeptical the more I've learned. I think it will arrive sometime and I eagerly await the day. But my early optimism (because I personally hate driving and definitely trust engineering solutions more than my fellow human drivers) is tempered because all the tests rely on some asterisk or another. Google cars, for example, require careful mapping of the territory (according to wiki inch-scale resolution) and slow down drastically when something unexpected comes up. I've driven past them and yeah, they do drive that way too. Evidence of safety problems caused by partial autonomy also muddy the actual path to a big rollout.
Basically my dream was to get a commute where I no longer have to drive. I no longer think that's happening much before I retire (say, a decade or so) and there hasn't been anything in this thread to change my mind. Not sure what you're seeing that makes you think it's right around the corner as opposed to niche thing.
posted by mark k at 8:42 PM on June 28, 2016 [2 favorites]
There are a lot of luddite denialists who think self-driving vehicles are "always ten years away" like fusion reactors.
Autonomous cars for consumers are easily a decade away. Probably two. The legal aspects of liability will likely delay it for 5 years on its own. Autonomous cars are coming (and not soon enough for my liking, to replace all the idiots who text/drink and otherwise don't actually *drive* when they are in their cars) but consumer-available self-driving cars that would be usable on a personal scale? Many, many years away.
The first stage is, I suspect, would be autonomous buses. On known, fixed, highly repetitious routes. They have started now (on a tiny and not very useful scale) as mentioned above, but even that being widespread is easily 5 years away. Likely even ten. Things of this complexity just do not move that fast and the technology is in its infancy, albeit growing fast.
There is a very, very big difference between saying 'autonomous vehicle technology is vaporware or will never have any impact on society' and the very real perspective that the automotive industry works on a 7-10 year turnover for a *model change*. The idea that it will be anything less than two model cycles to go from entirely manual to full automation is ludicrous. That's even without the massive legislative and liability issue of how the automation would work in interaction with other people (ie, what takes precedence, car occupant safety? Pedestrian safety? Bike safety OR Pedestrian? Other vehicle occupant safety?).
posted by Brockles at 9:18 PM on June 28, 2016
Autonomous cars for consumers are easily a decade away. Probably two. The legal aspects of liability will likely delay it for 5 years on its own. Autonomous cars are coming (and not soon enough for my liking, to replace all the idiots who text/drink and otherwise don't actually *drive* when they are in their cars) but consumer-available self-driving cars that would be usable on a personal scale? Many, many years away.
The first stage is, I suspect, would be autonomous buses. On known, fixed, highly repetitious routes. They have started now (on a tiny and not very useful scale) as mentioned above, but even that being widespread is easily 5 years away. Likely even ten. Things of this complexity just do not move that fast and the technology is in its infancy, albeit growing fast.
There is a very, very big difference between saying 'autonomous vehicle technology is vaporware or will never have any impact on society' and the very real perspective that the automotive industry works on a 7-10 year turnover for a *model change*. The idea that it will be anything less than two model cycles to go from entirely manual to full automation is ludicrous. That's even without the massive legislative and liability issue of how the automation would work in interaction with other people (ie, what takes precedence, car occupant safety? Pedestrian safety? Bike safety OR Pedestrian? Other vehicle occupant safety?).
posted by Brockles at 9:18 PM on June 28, 2016
mark k—If you can look at a bus route serviced by a self-driving vehicle running on normal roads mixed with other traffic and not only admit it actually exists, but also not try to pretend that no one would want it in the first place, your position isn't what I'm calling denialism.
Like, in the first comment in this thread you appear to simply be expressing disappointment at the speed at which autonomous cars will become widely available for personal use. We probably have pretty similar perceptions of this. I don't think autonomous vehicles have to replace every conventional vehicle on Earth, or be usable everywhere, or go really fast, or initially find optimal applications most suitable for the mainstream consumer, or even initially be most widely used in stable Western democracies, to have a substantial effect on our societies and economies; so maybe you would call me a skeptic too.
posted by XMLicious at 9:29 PM on June 28, 2016
Like, in the first comment in this thread you appear to simply be expressing disappointment at the speed at which autonomous cars will become widely available for personal use. We probably have pretty similar perceptions of this. I don't think autonomous vehicles have to replace every conventional vehicle on Earth, or be usable everywhere, or go really fast, or initially find optimal applications most suitable for the mainstream consumer, or even initially be most widely used in stable Western democracies, to have a substantial effect on our societies and economies; so maybe you would call me a skeptic too.
posted by XMLicious at 9:29 PM on June 28, 2016
Autonomous cars for consumers are easily a decade away. Probably two. The legal aspects of liability will likely delay it for 5 years on its own.
That was true maybe three years ago when Elon Musk was swearing up and down he wasn't working on autonomous vehicles, until this year where he admits he totally was. The first autonomous big rigs traveled the breadth of the EU just this year, and the drivers got to goof off the whole way.
2018, the latest, offered as an option as a luxury feature on a luxury model from a luxury make.
Bone stock standard on all 2024 vehicles.
posted by Slap*Happy at 9:43 PM on June 28, 2016
That was true maybe three years ago when Elon Musk was swearing up and down he wasn't working on autonomous vehicles, until this year where he admits he totally was. The first autonomous big rigs traveled the breadth of the EU just this year, and the drivers got to goof off the whole way.
2018, the latest, offered as an option as a luxury feature on a luxury model from a luxury make.
Bone stock standard on all 2024 vehicles.
posted by Slap*Happy at 9:43 PM on June 28, 2016
Not a hope. There simply isn't the product cycle time to produce that. Besides, the industry will never make it standard until it is mandatory - far, FAR too much cost to give away for nothing.
So, I'll take your bet. I think advanced cruise control/auto pilot (which, I may add, is not even an option on most luxury cars in any meaningful sense right now, yet) may be considered a 'can't be without it' offered COST option by 2024 on luxury models (So, 5/7 BMW, Merc S etc and above) and maybe dribbling down to real people cars as a horrifically expensive option by 2026. NO chance of it being standard before 2030. None.
There is no way, no how, the auto industry will be able to develop, produce, ammortise and profit from autonomy in enough volume to make it standard on even flagship models in 8 years. Hell, half of them will be at end of life of their 2018 model cycle at that stage so it would be impossible to integrate, being as the 2018 model years were largely designed two years ago. To reitterate, the cars that will be filling the showrooms in 2018 have already been designed. Autonomous cars are still a (realistic but still) pipe dream.
Car company's move at a glacial pace compared to start-up/niche manufacturers like Tesla (and.... ok, just Tesla). Even Tesla's customer ready cruise control is not legal to use in any autonomous sense, and is technically limited to highway use. It will be many, many years before legislators allow its use on even most roads, and major manufacturers just will not develop something if they can't sell it to the masses.
Tesla will lead the way, but they won't spearhead a shallow arrow of development. They will be 5 years ahead of major manufacturer because the car industry will wait until Tesla falls on its arse or not before committing dev funding to it in any meaningful way beyond skunk works stuff.
posted by Brockles at 10:02 PM on June 28, 2016
So, I'll take your bet. I think advanced cruise control/auto pilot (which, I may add, is not even an option on most luxury cars in any meaningful sense right now, yet) may be considered a 'can't be without it' offered COST option by 2024 on luxury models (So, 5/7 BMW, Merc S etc and above) and maybe dribbling down to real people cars as a horrifically expensive option by 2026. NO chance of it being standard before 2030. None.
There is no way, no how, the auto industry will be able to develop, produce, ammortise and profit from autonomy in enough volume to make it standard on even flagship models in 8 years. Hell, half of them will be at end of life of their 2018 model cycle at that stage so it would be impossible to integrate, being as the 2018 model years were largely designed two years ago. To reitterate, the cars that will be filling the showrooms in 2018 have already been designed. Autonomous cars are still a (realistic but still) pipe dream.
Car company's move at a glacial pace compared to start-up/niche manufacturers like Tesla (and.... ok, just Tesla). Even Tesla's customer ready cruise control is not legal to use in any autonomous sense, and is technically limited to highway use. It will be many, many years before legislators allow its use on even most roads, and major manufacturers just will not develop something if they can't sell it to the masses.
Tesla will lead the way, but they won't spearhead a shallow arrow of development. They will be 5 years ahead of major manufacturer because the car industry will wait until Tesla falls on its arse or not before committing dev funding to it in any meaningful way beyond skunk works stuff.
posted by Brockles at 10:02 PM on June 28, 2016
So, I'll take your bet. I think advanced cruise control/auto pilot (which, I may add, is not even an option on most luxury cars in any meaningful sense right now, yet) may be considered a 'can't be without it' offered COST option by 2024 on luxury models (So, 5/7 BMW, Merc S etc and above) and maybe dribbling down to real people cars as a horrifically expensive option by 2026. NO chance of it being standard before 2030. None.
I'm fairly sure "Distronic Plus" (the Mercedes name for the Mobileye system Tesla has branded "Auto Pilot") is already standard on S class. The BMW 7 series has its own version as standard too.
You can option a similar camera based system on most moderately expensive cars released in the last year or two from other manufacturers like Subaru, Hyundai, Mazda, Audi, Lexus and Infiniti.
posted by zymil at 12:31 AM on June 29, 2016
I'm fairly sure "Distronic Plus" (the Mercedes name for the Mobileye system Tesla has branded "Auto Pilot") is already standard on S class. The BMW 7 series has its own version as standard too.
You can option a similar camera based system on most moderately expensive cars released in the last year or two from other manufacturers like Subaru, Hyundai, Mazda, Audi, Lexus and Infiniti.
posted by zymil at 12:31 AM on June 29, 2016
You linked to video of a self-driving bus that moves at "up to" 12 mph--that is, a vehicle that can be outpaced by a long-distance runner
How long before I can get a long-distance runner to carry me to work?
posted by EndsOfInvention at 3:22 AM on June 29, 2016
How long before I can get a long-distance runner to carry me to work?
posted by EndsOfInvention at 3:22 AM on June 29, 2016
I'm a robot car fanboi but I'll chime in with one issue that may significantly slow adoption. Sensors.
Cost of sensors will not change as rapidly as software or microchips, the laser LIDAR ranging devices are a relatively new tech and transitioning to high volume economies of scale low cost devices will be a challenge.
Cost of sensor installation, maintenance and validation. A minor fender bender is not your fault but breaks a sensor and replacement is a few thousand and takes the car off the road for a few weeks? Not going to be acceptable.
How are the lenses of the cameras on cars kept clean? Will the care refuse to move until someone cleans all the cameras, even one that's hard to reach for a small old woman?
And a real issue is snow, all the sensors and tech is just not working in a snow storm or heavy rain.
There are lot's of real issues but just like I'm not going to take someones cell phone away for a day, just blink and most of the population will not even remember driving in a few years.
posted by sammyo at 4:33 AM on June 29, 2016
Cost of sensors will not change as rapidly as software or microchips, the laser LIDAR ranging devices are a relatively new tech and transitioning to high volume economies of scale low cost devices will be a challenge.
Cost of sensor installation, maintenance and validation. A minor fender bender is not your fault but breaks a sensor and replacement is a few thousand and takes the car off the road for a few weeks? Not going to be acceptable.
How are the lenses of the cameras on cars kept clean? Will the care refuse to move until someone cleans all the cameras, even one that's hard to reach for a small old woman?
And a real issue is snow, all the sensors and tech is just not working in a snow storm or heavy rain.
There are lot's of real issues but just like I'm not going to take someones cell phone away for a day, just blink and most of the population will not even remember driving in a few years.
posted by sammyo at 4:33 AM on June 29, 2016
Oh and one usability issue, somewhat unspoken by the tester but I've got the impression that the software is rather conservative (yes, insert a "duh") but which results in more breaking than human drivers and a bit of a jerky ride. That will get modified before a major beta test I'm sure.
posted by sammyo at 4:54 AM on June 29, 2016
posted by sammyo at 4:54 AM on June 29, 2016
Tesla will lead the way,
I'm not so sure about this. Just about every feature or option you can get on a modern car was first introduced on the Mercedes-Benz S-class. First car sold with fuel injection, seatbelt pre-tentioners, ABS brakes, turbo-diesel, side air bags, driver-side air bag, traction control, power windows, and precrash collision avoidance were all first available on the S-class Benz.
They're shifting a little bit next year and putting all the hot new tech on the smaller E-class including a system (really a combination of systems) up to 130 mph. It requires the driver's hands to be on the wheel but I wonder how much of that is due to legal liability issues and how much is because the engineers think their system needs a human backup to be paying attention at all times.
Audi has been talking about their new A4 and it's system that will drive autonomously up to 40mph in heavy traffic for a while. The Merc's system seems to be a bit more advanced and it only came out six months later.
I don't know how far away 100% autonomous vehicles really are but getting to nearly 100% autonomous driving on the highway, is less than five years away. And frankly, I don't care that much about my car being able to drive itself 100% of the time, just the ability to get on the interstate and turn on auto-pilot until my exit is coming up would be a paradigm shift.
posted by VTX at 5:43 AM on June 29, 2016 [1 favorite]
I'm not so sure about this. Just about every feature or option you can get on a modern car was first introduced on the Mercedes-Benz S-class. First car sold with fuel injection, seatbelt pre-tentioners, ABS brakes, turbo-diesel, side air bags, driver-side air bag, traction control, power windows, and precrash collision avoidance were all first available on the S-class Benz.
They're shifting a little bit next year and putting all the hot new tech on the smaller E-class including a system (really a combination of systems) up to 130 mph. It requires the driver's hands to be on the wheel but I wonder how much of that is due to legal liability issues and how much is because the engineers think their system needs a human backup to be paying attention at all times.
Audi has been talking about their new A4 and it's system that will drive autonomously up to 40mph in heavy traffic for a while. The Merc's system seems to be a bit more advanced and it only came out six months later.
I don't know how far away 100% autonomous vehicles really are but getting to nearly 100% autonomous driving on the highway, is less than five years away. And frankly, I don't care that much about my car being able to drive itself 100% of the time, just the ability to get on the interstate and turn on auto-pilot until my exit is coming up would be a paradigm shift.
posted by VTX at 5:43 AM on June 29, 2016 [1 favorite]
just the ability to get on the interstate and turn on auto-pilot until my exit is coming up would be a paradigm shift.
Agreed. Not least because it is the most boring part of a journey, so is more liable to distraction. That'd be nice.
Audi has been talking about their new A4 and it's system that will drive autonomously up to 40mph in heavy traffic for a while.
Yeah, but.... 'follow the other cars' is not exactly autonomous. That's basically just active cruise and lane control, which in an extremely limited parameter like heavy traffic, is relying on most of the other cars for its reference. The fact that it is specified as 'heavy traffic' and 'limited time' speaks volumes as to the limitations on the system.
Just about every feature or option you can get on a modern car was first introduced on the Mercedes-Benz S-class.
Yeah, good point. But Merc S don't 'lead' as such. They just quietly add stuff because the people who buy the S don't really look anywhere else. Getting the tech in front of the 'normals' seems to be something Tesla are at the front of. Not least in marketing.
posted by Brockles at 6:06 AM on June 29, 2016
Agreed. Not least because it is the most boring part of a journey, so is more liable to distraction. That'd be nice.
Audi has been talking about their new A4 and it's system that will drive autonomously up to 40mph in heavy traffic for a while.
Yeah, but.... 'follow the other cars' is not exactly autonomous. That's basically just active cruise and lane control, which in an extremely limited parameter like heavy traffic, is relying on most of the other cars for its reference. The fact that it is specified as 'heavy traffic' and 'limited time' speaks volumes as to the limitations on the system.
Just about every feature or option you can get on a modern car was first introduced on the Mercedes-Benz S-class.
Yeah, good point. But Merc S don't 'lead' as such. They just quietly add stuff because the people who buy the S don't really look anywhere else. Getting the tech in front of the 'normals' seems to be something Tesla are at the front of. Not least in marketing.
posted by Brockles at 6:06 AM on June 29, 2016
And a real issue is snow, all the sensors and tech is just not working in a snow storm or heavy rain.
Exactly. It is no coincidence that most of the testing is being done in one of the more stable weather climates the US has (same for the Greek bus). Anyone know if Google et al test outside Cali/Nevada? I haven't heard of it yet.
posted by Brockles at 6:08 AM on June 29, 2016
Exactly. It is no coincidence that most of the testing is being done in one of the more stable weather climates the US has (same for the Greek bus). Anyone know if Google et al test outside Cali/Nevada? I haven't heard of it yet.
posted by Brockles at 6:08 AM on June 29, 2016
Cost of sensors will not change as rapidly as software or microchips
One of my general complaints in technology discussions is that many people's--and especially the press's--intuitions are centered around the computing world. If you have a feature on a smart phone that sort of works today, the only reason it wouldn't be standard once chips are cheaper in two years is if it doesn't sell. With it's quick development times this is the one area someone in there mid-30s or so has seen many rounds of development and has had a chance to observe how tech works.
Inventions that aren't 100% digital but rely on things in the physical world working out are a whole different ball of wax. We don't have commercial supersonic airplanes, we haven't cured cancer with therapeutic antibodies or siRNA, Google trends can't predict the next flu season after all, and a successful fusion reactor design is still unknown. But many times they've been written about as if development is as straightforward as just saying "ten years development" instead of "two."
Self-driving cars have a large computing element so I originally binned them as digital tech. I don't think that's right. I now think the real challenges are around the physical world and people and there's no guarantee that translating from places where they work (freeways, 10-mph speeds on defined routes, places where every fire hydrant and traffic light is digitally mapped ahead of time) to conditions on my daily commute is straightforward.
Anyone know if Google et al test outside Cali/Nevada? I haven't heard of it yet.
Volvo was testing extensively on some Baltic Island--Gotland I think but would not swear to it.
Of course googling to jog my memory I found this. (Which to be fair is not an "autonomous" car failure, and the human driver was rather stupider than the microchip.)
posted by mark k at 6:48 AM on June 29, 2016
One of my general complaints in technology discussions is that many people's--and especially the press's--intuitions are centered around the computing world. If you have a feature on a smart phone that sort of works today, the only reason it wouldn't be standard once chips are cheaper in two years is if it doesn't sell. With it's quick development times this is the one area someone in there mid-30s or so has seen many rounds of development and has had a chance to observe how tech works.
Inventions that aren't 100% digital but rely on things in the physical world working out are a whole different ball of wax. We don't have commercial supersonic airplanes, we haven't cured cancer with therapeutic antibodies or siRNA, Google trends can't predict the next flu season after all, and a successful fusion reactor design is still unknown. But many times they've been written about as if development is as straightforward as just saying "ten years development" instead of "two."
Self-driving cars have a large computing element so I originally binned them as digital tech. I don't think that's right. I now think the real challenges are around the physical world and people and there's no guarantee that translating from places where they work (freeways, 10-mph speeds on defined routes, places where every fire hydrant and traffic light is digitally mapped ahead of time) to conditions on my daily commute is straightforward.
Anyone know if Google et al test outside Cali/Nevada? I haven't heard of it yet.
Volvo was testing extensively on some Baltic Island--Gotland I think but would not swear to it.
Of course googling to jog my memory I found this. (Which to be fair is not an "autonomous" car failure, and the human driver was rather stupider than the microchip.)
posted by mark k at 6:48 AM on June 29, 2016
Anyone know if Google et al test outside Cali/Nevada? I haven't heard of it yet.
They're testing here in Austin. Not exactly a lot of climate extremes beyond (this year and last year) a lot of rain. There are probably other places they're testing that are low-key enough that most people don't know about it.
posted by RustyBrooks at 7:47 AM on June 29, 2016
They're testing here in Austin. Not exactly a lot of climate extremes beyond (this year and last year) a lot of rain. There are probably other places they're testing that are low-key enough that most people don't know about it.
posted by RustyBrooks at 7:47 AM on June 29, 2016
And a real issue is snow, all the sensors and tech is just not working in a snow storm or heavy rain.
Big deal. Humans crash all the time in snow and heavy rain.
posted by EndsOfInvention at 7:51 AM on June 29, 2016
Big deal. Humans crash all the time in snow and heavy rain.
posted by EndsOfInvention at 7:51 AM on June 29, 2016
Sorry, that sounded dismissive. What I mean is, humans crash all the time in snow and rain (come watch UK drivers try and cope with 2cm of snow on the ground), and never get any better. An autonomous system can continually improve and get better over time. I would expect an autonomous car to at least be "as good" to begin with - sure it might crash, but like I said, that happens anyway.
posted by EndsOfInvention at 8:16 AM on June 29, 2016 [1 favorite]
posted by EndsOfInvention at 8:16 AM on June 29, 2016 [1 favorite]
Traveling in snow and ice will be tricky for cars that depend on visual sensors, just as we have trouble using our visual sensors when traveling in snow and ice. Driverless cars and human drivers could both use an inertial navigation system to help keep them on the road.
posted by pracowity at 9:24 AM on June 29, 2016
posted by pracowity at 9:24 AM on June 29, 2016
It's not about the capability of autonomous systems to drive in the snow because of the road conditions, it's because the general wet, cold, icy, grime that cakes the roads all winter will cover the sensors and the car won't be able to see.
Yeah, good point. But Merc S don't 'lead' as such. They just quietly add stuff because the people who buy the S don't really look anywhere else.
I think it's just a more reserved version of taking the lead. When I was sold cars for a living, we would always read about the new S-class because whatever new features and technology it had was stuff we could count on starting to show up on our cars (Nissan's usually) a few years down the line. I think they lead the way, they're just really quite about it. The switch to launching the new tech on the E-class is, I think, a way of making their leadership role a little more visible. But then it's also a reaction to Audi's A4 and Tesla.
Tesla has stood up and said, "We will lead the way forward with technology!" Mercedes's head snapped up when they heard that so they said, "Wait just a minute, we've been pushing the technology envelope for 50 years! We've been here literally the whole time!"
So maybe Mercedes is just catching up but maybe Tesla just makes a bunch of noise (it seems like every story of their next big thing is followed by a story about the problems with that thing) and they've poked the quiet guy in the corner so now they'll find out what real advancements look like.
I've long ago lost any sense of brand loyalty, especially when it comes to cars so it doesn't really matter which of is more right (neither of us is wrong) because in the end, it just gets me my interstate auto-pilot faster and that's the part I really care about. Even though I think the Merc looks better than the Tesla, mostly because the styling isn't as "loud".
posted by VTX at 10:08 AM on June 29, 2016
Yeah, good point. But Merc S don't 'lead' as such. They just quietly add stuff because the people who buy the S don't really look anywhere else.
I think it's just a more reserved version of taking the lead. When I was sold cars for a living, we would always read about the new S-class because whatever new features and technology it had was stuff we could count on starting to show up on our cars (Nissan's usually) a few years down the line. I think they lead the way, they're just really quite about it. The switch to launching the new tech on the E-class is, I think, a way of making their leadership role a little more visible. But then it's also a reaction to Audi's A4 and Tesla.
Tesla has stood up and said, "We will lead the way forward with technology!" Mercedes's head snapped up when they heard that so they said, "Wait just a minute, we've been pushing the technology envelope for 50 years! We've been here literally the whole time!"
So maybe Mercedes is just catching up but maybe Tesla just makes a bunch of noise (it seems like every story of their next big thing is followed by a story about the problems with that thing) and they've poked the quiet guy in the corner so now they'll find out what real advancements look like.
I've long ago lost any sense of brand loyalty, especially when it comes to cars so it doesn't really matter which of is more right (neither of us is wrong) because in the end, it just gets me my interstate auto-pilot faster and that's the part I really care about. Even though I think the Merc looks better than the Tesla, mostly because the styling isn't as "loud".
posted by VTX at 10:08 AM on June 29, 2016
It's not about the capability of autonomous systems to drive in the snow because of the road conditions, it's because the general wet, cold, icy, grime that cakes the roads all winter will cover the sensors and the car won't be able to see.
It is also about the road conditions, though. There can be very rapidly changing road quality variations even in the dry in terms of camber, friction, external factors (water, ice, detritus, road kill, shadow/sunlight, moss/mud/dust etc). There is a vast chasm in capabilities between a car dealing with a closely mapped environment (which is how I understand Google does things) and a truly autonomous car that can interpret new surroundings as it goes PLUS cope with things like ice still under a shaded area, or on a bridge etc. Or a dry road that still has a stream across it around a corner because the fields either side are still flooding.
And if this is all done on visual sensors, then it will need to be constantly cleaned (presumably automatically) to even function, let alone then cope with the conditions that produced the dirt/snow etc.
posted by Brockles at 10:46 AM on June 29, 2016
It is also about the road conditions, though. There can be very rapidly changing road quality variations even in the dry in terms of camber, friction, external factors (water, ice, detritus, road kill, shadow/sunlight, moss/mud/dust etc). There is a vast chasm in capabilities between a car dealing with a closely mapped environment (which is how I understand Google does things) and a truly autonomous car that can interpret new surroundings as it goes PLUS cope with things like ice still under a shaded area, or on a bridge etc. Or a dry road that still has a stream across it around a corner because the fields either side are still flooding.
And if this is all done on visual sensors, then it will need to be constantly cleaned (presumably automatically) to even function, let alone then cope with the conditions that produced the dirt/snow etc.
posted by Brockles at 10:46 AM on June 29, 2016
I thought the traction control and ABS systems would be able to cope with traction. Dynamic stability control can already steer a skidding car with the brakes, throttle and AWD systems. BMW's system is even a little bit predictive with their x-drive AWD system.
I would think that if you add the ability to control the steering more directly you could get an autonomous car to out drive a finish rally driver in the snow.
Then you're back to the "normal" issues of an autonomous car navigating the real world differently than it can on a race track (other cars, people, route variations, etc.). Regaining control after a little under-steer should be one of those problems that's easy for a computer but hard for a human.
I mean, Brockles:car threads::Physicsmatt:physics threads so I'll take your word for it but does a car really need LIDAR (and other sensors that snow could mess with) to drive in the snow?
posted by VTX at 11:00 AM on June 29, 2016
I would think that if you add the ability to control the steering more directly you could get an autonomous car to out drive a finish rally driver in the snow.
Then you're back to the "normal" issues of an autonomous car navigating the real world differently than it can on a race track (other cars, people, route variations, etc.). Regaining control after a little under-steer should be one of those problems that's easy for a computer but hard for a human.
I mean, Brockles:car threads::Physicsmatt:physics threads so I'll take your word for it but does a car really need LIDAR (and other sensors that snow could mess with) to drive in the snow?
posted by VTX at 11:00 AM on June 29, 2016
Brockles: "The first stage is, I suspect, would be autonomous buses. On known, fixed, highly repetitious routes. They have started now (on a tiny and not very useful scale) as mentioned above, but even that being widespread is easily 5 years away. Likely even ten. Things of this complexity just do not move that fast and the technology is in its infancy, albeit growing fast. "
We have the ideal test case for self driving vehicles here. The Vancouver to Kamloops trucking route can be done completely on limited access highway; most of it is controlled even for animals. Currently there are dozens of drivers driving back and forth every night. Put the depots right at an interchange and exposure to most of the liability inducing risks goes way down. And tractors have the advantage of having lots of room for equipment without much in the way of packaging concerns; already being partial hand build enabling easy prototyping and having regular monitoring and maintenance. If they can develop a tractor to save 100K in total compensation per truck every year in this sort of situation then I'll starting looking for autonomous cars.
Of course they'll have to deal with challenging weather conditions; that route has everything but cold below -40C and large hail.
VTX: "I'm not so sure about this. Just about every feature or option you can get on a modern car was first introduced on the Mercedes-Benz S-class. First car sold with fuel injection, seatbelt pre-tentioners, ABS brakes, turbo-diesel, side air bags, driver-side air bag, traction control, power windows, and precrash collision avoidance were all first available on the S-class Benz."
Electronic ABS was first used on the rear of 70 Lincolns and at all four wheels on 71 Imperials (Jensen was the first production car with any sort of ABS but it was mechanical and apparently didn't work all that well). Packard had the first power windows in 1940 before Mercedes even started building the S class. GM had airbags in the early 70s well ahead of Mercedes (though they were later discontinued by bean counting). Buick was the first with electronic traction control (mechanical means predating electronics) in '71 beating the S Class by a decade.
posted by Mitheral at 11:29 AM on June 29, 2016
We have the ideal test case for self driving vehicles here. The Vancouver to Kamloops trucking route can be done completely on limited access highway; most of it is controlled even for animals. Currently there are dozens of drivers driving back and forth every night. Put the depots right at an interchange and exposure to most of the liability inducing risks goes way down. And tractors have the advantage of having lots of room for equipment without much in the way of packaging concerns; already being partial hand build enabling easy prototyping and having regular monitoring and maintenance. If they can develop a tractor to save 100K in total compensation per truck every year in this sort of situation then I'll starting looking for autonomous cars.
Of course they'll have to deal with challenging weather conditions; that route has everything but cold below -40C and large hail.
VTX: "I'm not so sure about this. Just about every feature or option you can get on a modern car was first introduced on the Mercedes-Benz S-class. First car sold with fuel injection, seatbelt pre-tentioners, ABS brakes, turbo-diesel, side air bags, driver-side air bag, traction control, power windows, and precrash collision avoidance were all first available on the S-class Benz."
Electronic ABS was first used on the rear of 70 Lincolns and at all four wheels on 71 Imperials (Jensen was the first production car with any sort of ABS but it was mechanical and apparently didn't work all that well). Packard had the first power windows in 1940 before Mercedes even started building the S class. GM had airbags in the early 70s well ahead of Mercedes (though they were later discontinued by bean counting). Buick was the first with electronic traction control (mechanical means predating electronics) in '71 beating the S Class by a decade.
posted by Mitheral at 11:29 AM on June 29, 2016
I think that airbag was made from a Mercedes patent. Benz re-introduced the option in 1980 (presumably after they fixed the design).
They also invented the car.
And in a lot of the cases where they weren't the first, they were the first to get it work worth a damn. I'm not really sure what your point is. Are you trying to say that Mercedes-Benz is not, in fact, a technology leader in the auto industry? Or that I did not actually research the new E-class to find out what options where coming down the pipe?
posted by VTX at 12:42 PM on June 29, 2016
They also invented the car.
And in a lot of the cases where they weren't the first, they were the first to get it work worth a damn. I'm not really sure what your point is. Are you trying to say that Mercedes-Benz is not, in fact, a technology leader in the auto industry? Or that I did not actually research the new E-class to find out what options where coming down the pipe?
posted by VTX at 12:42 PM on June 29, 2016
does a car really need LIDAR (and other sensors that snow could mess with) to drive in the snow?
I don't see how it can do the rougher elements of control without some very precise knowledge of location and orientation. The abs/tcs/stability control stuff is all very well, but that is fine tuning of a course a human is still telling the car which way to go - all those systems do their best (and are excellent currently) at keeping as best as possible to the intended line, but there is still a lot of interpretation in how to modify that line once traction/grip has been lost. The car can have lost lateral position by some margin or its heading in the process of retaining 4 points of traction. What does it do then? Does it head to a place of safety? Or try and regain the original path as soon as possible? What about other traffic that it may now be heading towards or at different speeds/trajectories to? There are a lot of variables and factors that need to be considered.
Also, anticipation - how does a car notice that there is a tree or a great big puddle of mud/cow shit on the road? That is a significant enough grip level event that a human will see it early enough (barring incompetency) and brake in advance. The auto car can't necessarily do that, and so it has to deal with a sudden and very severe grip loss, which it will be unable to cope with. The process of stability control involves braking and changing yaw/angle using the other points of grip as a reference. If you get too far past that loss of grip (so, a >90 degree spin, or loss of grip at more than two points) then what is the car to do? Slam on the brakes and wait? Keep fighting the grip? Try and retain the original path? It would need to know what it is heading for, how that relates to its survivability protocols (ie save the passengers at all costs, or save the greatest number of people at all costs or sacrifice the passengers for the surroundings) and that is a totally untouched part of autonomous cars. As soon as the car goes more than 25-30% (say) away from its intended path, current tech just says "Oh shit, you're on your own, sunshine" and the bells and whistles turn off or become ineffective. They are not a full vehicle control system, they are just a 'help you along' system. It's not like you can just add a GPS controller onto the current car and let it go. It is far, far more complex than that. Especially when you bring ethical decisions into it - sacrifice the car or the surroundings? Head for the wall and trust the airbags as the fastest and safest way to immediately stop the car? Or head through the fence into the field and hope the control mechanisms can stop the car on grass/miss the cows... The possibilities need to be considered. There is much, much more to it than the current grip-maximising systems.
posted by Brockles at 1:47 PM on June 29, 2016 [1 favorite]
I don't see how it can do the rougher elements of control without some very precise knowledge of location and orientation. The abs/tcs/stability control stuff is all very well, but that is fine tuning of a course a human is still telling the car which way to go - all those systems do their best (and are excellent currently) at keeping as best as possible to the intended line, but there is still a lot of interpretation in how to modify that line once traction/grip has been lost. The car can have lost lateral position by some margin or its heading in the process of retaining 4 points of traction. What does it do then? Does it head to a place of safety? Or try and regain the original path as soon as possible? What about other traffic that it may now be heading towards or at different speeds/trajectories to? There are a lot of variables and factors that need to be considered.
Also, anticipation - how does a car notice that there is a tree or a great big puddle of mud/cow shit on the road? That is a significant enough grip level event that a human will see it early enough (barring incompetency) and brake in advance. The auto car can't necessarily do that, and so it has to deal with a sudden and very severe grip loss, which it will be unable to cope with. The process of stability control involves braking and changing yaw/angle using the other points of grip as a reference. If you get too far past that loss of grip (so, a >90 degree spin, or loss of grip at more than two points) then what is the car to do? Slam on the brakes and wait? Keep fighting the grip? Try and retain the original path? It would need to know what it is heading for, how that relates to its survivability protocols (ie save the passengers at all costs, or save the greatest number of people at all costs or sacrifice the passengers for the surroundings) and that is a totally untouched part of autonomous cars. As soon as the car goes more than 25-30% (say) away from its intended path, current tech just says "Oh shit, you're on your own, sunshine" and the bells and whistles turn off or become ineffective. They are not a full vehicle control system, they are just a 'help you along' system. It's not like you can just add a GPS controller onto the current car and let it go. It is far, far more complex than that. Especially when you bring ethical decisions into it - sacrifice the car or the surroundings? Head for the wall and trust the airbags as the fastest and safest way to immediately stop the car? Or head through the fence into the field and hope the control mechanisms can stop the car on grass/miss the cows... The possibilities need to be considered. There is much, much more to it than the current grip-maximising systems.
posted by Brockles at 1:47 PM on June 29, 2016 [1 favorite]
I would think that if you add the ability to control the steering more directly you could get an autonomous car to out drive a finish rally driver in the snow.
Absolutely not. There are huge, HUGE other parts that you aren't considering. Not least the fine art of controlled slides and anticipation involved. Plus the safest way is almost never the fastest way on snow, and often involves using terrain (ie ditches) to turn the car with - hooking the front wheel into a ditch and flooring the throttle to stay in the radius is a very effective method, but you need to allow for how much that ditch/hole/groove prevails through the corner and it changes depending on how many cars have been through the area before you.
On traction control, I got hold of a new Ford Focus with stability control a few weeks ago and 'happened' to be on some windy, smooth, unpaved roads in the desert with zero traffic. Good god, but that is an impressive system. Trust me in that I was being *seriously* abusive to the available grip and the car was phenomenally capable of getting the car back into the given line. I started off trying to play with the Scandinavian flick (pendulum movement of the rear of the car to change direction at a higher speed) and the damn car killed it EVERY time. Big cloud of dust and some unpleasant lurching, but extremely effective. No matter where I got the car (very VERY sideways and pretty fast) it got the car straight wherever I was pointing the wheel. It killed any fun in the car, sadly, but was technically impressive.
However, I realise now I think of it in this context that I changed my mind about which particular road/track I was going to go down a couple of times because of how long the car took to get everything tidied up. It was an area that would be a sub division, so lots of 'soon to be roads' and it was easy to say "Well, that isn't coming back, I'll just head this way instead". That's not a judgment call a computer can make. Would it know to use both sides of the road to recover its composure? Wouldn't it have to check if it had room/time to do so if it needed it? Or would it choose to crash if it needed the opposite lane for one end of the other of the car in the process of regaining stability?
It's not an easy answer.
posted by Brockles at 1:58 PM on June 29, 2016 [2 favorites]
Absolutely not. There are huge, HUGE other parts that you aren't considering. Not least the fine art of controlled slides and anticipation involved. Plus the safest way is almost never the fastest way on snow, and often involves using terrain (ie ditches) to turn the car with - hooking the front wheel into a ditch and flooring the throttle to stay in the radius is a very effective method, but you need to allow for how much that ditch/hole/groove prevails through the corner and it changes depending on how many cars have been through the area before you.
On traction control, I got hold of a new Ford Focus with stability control a few weeks ago and 'happened' to be on some windy, smooth, unpaved roads in the desert with zero traffic. Good god, but that is an impressive system. Trust me in that I was being *seriously* abusive to the available grip and the car was phenomenally capable of getting the car back into the given line. I started off trying to play with the Scandinavian flick (pendulum movement of the rear of the car to change direction at a higher speed) and the damn car killed it EVERY time. Big cloud of dust and some unpleasant lurching, but extremely effective. No matter where I got the car (very VERY sideways and pretty fast) it got the car straight wherever I was pointing the wheel. It killed any fun in the car, sadly, but was technically impressive.
However, I realise now I think of it in this context that I changed my mind about which particular road/track I was going to go down a couple of times because of how long the car took to get everything tidied up. It was an area that would be a sub division, so lots of 'soon to be roads' and it was easy to say "Well, that isn't coming back, I'll just head this way instead". That's not a judgment call a computer can make. Would it know to use both sides of the road to recover its composure? Wouldn't it have to check if it had room/time to do so if it needed it? Or would it choose to crash if it needed the opposite lane for one end of the other of the car in the process of regaining stability?
It's not an easy answer.
posted by Brockles at 1:58 PM on June 29, 2016 [2 favorites]
Okay, so I went back and looked at how we got on this topic. I was saying that the problem with snow isn't so much that it's slippery, but that the sensors will get caked with grime and the car won't be able to "see". So the LIDAR bit was the wrong direction. So assume a car can see and all the sensors can deal with the dirt, grime, snow, etc. In that case, I think there is a lot that an autonomous car can compensate for.
The example you gave about the Ford* is a good one though. If taking the intended route is no longer an option, the car now is now forced to make a decision. In some cases, I would suppose that the car would be able to pull itself over and ask it's human what to do. It would super easy if all that happened is that the car starts braking for a turn at an uncontrolled intersection and hits some ice then slides right past. Until that problem gets solved, a car can't drive itself without a human present. As long as you can get the car to safely get itself to the shoulder, napping should still be okay, otherwise you've got to be ready to take the wheel at all times.
*I think the RS version can be put into a drift mode that will let you get more wild before shutting things down and tries to be smoother about it when it does.
posted by VTX at 2:50 PM on June 29, 2016
The example you gave about the Ford* is a good one though. If taking the intended route is no longer an option, the car now is now forced to make a decision. In some cases, I would suppose that the car would be able to pull itself over and ask it's human what to do. It would super easy if all that happened is that the car starts braking for a turn at an uncontrolled intersection and hits some ice then slides right past. Until that problem gets solved, a car can't drive itself without a human present. As long as you can get the car to safely get itself to the shoulder, napping should still be okay, otherwise you've got to be ready to take the wheel at all times.
*I think the RS version can be put into a drift mode that will let you get more wild before shutting things down and tries to be smoother about it when it does.
posted by VTX at 2:50 PM on June 29, 2016
VTX: "I'm not really sure what your point is."
No point per se. Just hate to see incorrect information on this topic spread on the internet.
posted by Mitheral at 6:00 PM on June 29, 2016
No point per se. Just hate to see incorrect information on this topic spread on the internet.
posted by Mitheral at 6:00 PM on June 29, 2016
Also, anticipation - how does a car notice that there is a tree or a great big puddle of mud/cow shit on the road?
This is the really cool part that's making this all possible - because the goddamn computer software has seen it before, and can judge the likelihood of it being a tree or great big puddle of mud or of cow-shit or just a shadow. It will then anticipate driving strategies, including hundreds of alternate ones it can switch to as it senses things coming out of whack.
Machine learning is a very interesting field at the moment. Up until Google thought, "Hey! Self driving cars!" it was used mostly for marketing research... so, yes, it works.
Here's how it goes. You feed as much info as you can to "Big-Data" tools, and then write software to analyze them for trends and patterns, and then write software to apply those trends and patterns to machine vision systems and xyz-axis accelerometers and servos and control instrumentation. Yes, the computer will know the car's ass-end is squirrelly, and yes the computer will notice the oncoming giant tourbus incapable of staying on its side of the double-yellows, and come up with a drive solution that will put you in line in time, or maybe go off the shoulder a little bit, where the sudden rough going under its curb-side wheel is also something it's experienced before, and has seen dealt with by a human, and has been analyzed and applied to its sensor and control systems.
In short, they put a computer into a very nice car with someone like you at the wheel, with sensors recording your every move on the throttle, brake and wheel, with all of the vision and motion sensors adding to the pile of data, and have you drive for a few million miles in it.
The computers in the current gen self-drivers weren't programmed to drive. They learned to drive, their hand at the wheel, their foot at the pedals right alongside yours as they watched with eyes, ears and a sense of motion better than yours will ever be, attent and learning what you teach with your actual driving. Millions of miles, never getting bored, just learning what to do, when to do it, from excellent drivers in all conditions.
What about bad drivers who Google hired thinking they were good? The tools recognize bad data and exclude it from the model. And, at some point, you have to trust in the team of people in charge of data acquisition to do their jobs and well.
Programmers then go in and set limits and priorities, because, yeah? Who wants Skynet in a Subaru? And I know that at least Google actually has real-live ethicists on staff for this. And by Google, I mean Alphabet, but it started as Google, as did Alphabet.
And self-drivers are being tested in CA, NV, NJ and VA in the US, and self-driver big rigs from Volvo have been prowling Gotland, part of Sweden in the cold part of the Baltic, for years now.
Snow and ice in all of their varieties isn't a problem. It simply isn't.
There are too many hits, and a lot of them good - google or bing or whatever "self driving cars machine learning" - one kid in his 20's made a viable system that Elon Musk rejected, as his own team had been going that way already and had more sophisticated data models to work with.
posted by Slap*Happy at 6:50 PM on June 29, 2016
This is the really cool part that's making this all possible - because the goddamn computer software has seen it before, and can judge the likelihood of it being a tree or great big puddle of mud or of cow-shit or just a shadow. It will then anticipate driving strategies, including hundreds of alternate ones it can switch to as it senses things coming out of whack.
Machine learning is a very interesting field at the moment. Up until Google thought, "Hey! Self driving cars!" it was used mostly for marketing research... so, yes, it works.
Here's how it goes. You feed as much info as you can to "Big-Data" tools, and then write software to analyze them for trends and patterns, and then write software to apply those trends and patterns to machine vision systems and xyz-axis accelerometers and servos and control instrumentation. Yes, the computer will know the car's ass-end is squirrelly, and yes the computer will notice the oncoming giant tourbus incapable of staying on its side of the double-yellows, and come up with a drive solution that will put you in line in time, or maybe go off the shoulder a little bit, where the sudden rough going under its curb-side wheel is also something it's experienced before, and has seen dealt with by a human, and has been analyzed and applied to its sensor and control systems.
In short, they put a computer into a very nice car with someone like you at the wheel, with sensors recording your every move on the throttle, brake and wheel, with all of the vision and motion sensors adding to the pile of data, and have you drive for a few million miles in it.
The computers in the current gen self-drivers weren't programmed to drive. They learned to drive, their hand at the wheel, their foot at the pedals right alongside yours as they watched with eyes, ears and a sense of motion better than yours will ever be, attent and learning what you teach with your actual driving. Millions of miles, never getting bored, just learning what to do, when to do it, from excellent drivers in all conditions.
What about bad drivers who Google hired thinking they were good? The tools recognize bad data and exclude it from the model. And, at some point, you have to trust in the team of people in charge of data acquisition to do their jobs and well.
Programmers then go in and set limits and priorities, because, yeah? Who wants Skynet in a Subaru? And I know that at least Google actually has real-live ethicists on staff for this. And by Google, I mean Alphabet, but it started as Google, as did Alphabet.
And self-drivers are being tested in CA, NV, NJ and VA in the US, and self-driver big rigs from Volvo have been prowling Gotland, part of Sweden in the cold part of the Baltic, for years now.
Snow and ice in all of their varieties isn't a problem. It simply isn't.
There are too many hits, and a lot of them good - google or bing or whatever "self driving cars machine learning" - one kid in his 20's made a viable system that Elon Musk rejected, as his own team had been going that way already and had more sophisticated data models to work with.
posted by Slap*Happy at 6:50 PM on June 29, 2016
The coolest thing about driverless cars will be when they are linked, whether directly or through the same remote databases. Only one car will ever hit that new pothole, because every car that follows it will have been warned by the car that lost a hub cap on it. And news of a serious accident will immediately ripple through every other car for 10 miles around, so everyone is cautious approaching the accident, gets out of the way for (driverless) emergency vehicles racing up the cleared fast lane (or middle of the road), and starts looking for alternative routes. The roads will be more like a circulatory system.
posted by pracowity at 12:31 AM on June 30, 2016 [1 favorite]
posted by pracowity at 12:31 AM on June 30, 2016 [1 favorite]
Only one car will ever hit that new pothole, because every car that follows it will have been warned by the car that lost a hub cap on it.
As long as you've subscribed to Road Updates Premium, of course (just 49.99/month!).
posted by EndsOfInvention at 6:28 AM on June 30, 2016
As long as you've subscribed to Road Updates Premium, of course (just 49.99/month!).
posted by EndsOfInvention at 6:28 AM on June 30, 2016
As long as you've subscribed to Road Updates Premium, of course (just 49.99/month!).
Right. The true value of autonomous cars will be in the hive mentality, which requires cars of different types and manufacturers to communicate. Which will require legislation, and until then it will be a premium only thing. Which will take ages of fighting about.
Snow and ice in all of their varieties isn't a problem. It simply isn't.
That's a compelling argument you are totally failing to make there. Wait, not it isn't. Just say it is so....
Slippery surfaces are the hardest thing to account for even right now. TCS and stability systems are in constant flux and development. We can't deal with them reliably even WITH a human driver. It's just not feasible at the current level of tech to just *pop* the human computer out of that mix. The systems just aren't capable enough even without the direct control and decision making being in place.
Millions of miles, never getting bored, just learning what to do, when to do it, from excellent drivers in all conditions.
Yet this is somehow achievable to collect, analyse and utilise all that data AND feeding that back into a working autonomous car all within a couple of years? How long will it take enough people to drive enough distance to get those millions of miles? Quite a while, surely?
posted by Brockles at 4:09 PM on June 30, 2016 [1 favorite]
Right. The true value of autonomous cars will be in the hive mentality, which requires cars of different types and manufacturers to communicate. Which will require legislation, and until then it will be a premium only thing. Which will take ages of fighting about.
Snow and ice in all of their varieties isn't a problem. It simply isn't.
That's a compelling argument you are totally failing to make there. Wait, not it isn't. Just say it is so....
Slippery surfaces are the hardest thing to account for even right now. TCS and stability systems are in constant flux and development. We can't deal with them reliably even WITH a human driver. It's just not feasible at the current level of tech to just *pop* the human computer out of that mix. The systems just aren't capable enough even without the direct control and decision making being in place.
Millions of miles, never getting bored, just learning what to do, when to do it, from excellent drivers in all conditions.
Yet this is somehow achievable to collect, analyse and utilise all that data AND feeding that back into a working autonomous car all within a couple of years? How long will it take enough people to drive enough distance to get those millions of miles? Quite a while, surely?
posted by Brockles at 4:09 PM on June 30, 2016 [1 favorite]
Well, here we are. The first death due to a self-driving car crash (a Tesla in "auto-pilot" mode - not fully autonomous I think?).
Guardian article
Tesla statement
posted by EndsOfInvention at 4:13 PM on June 30, 2016
Guardian article
Tesla statement
posted by EndsOfInvention at 4:13 PM on June 30, 2016
Interesting description of the liability avoidance measures of the Tesla AutoPilot system:
posted by Mitheral at 5:01 PM on June 30, 2016
When drivers activate Autopilot, the acknowledgment box explains, among other things, that Autopilot “is an assist feature that requires you to keep your hands on the steering wheel at all times," and that "you need to maintain control and responsibility for your vehicle” while using it. Additionally, every time that Autopilot is engaged, the car reminds the driver to “Always keep your hands on the wheel. Be prepared to take over at any time.” The system also makes frequent checks to ensure that the driver's hands remain on the wheel and provides visual and audible alerts if hands-on is not detected. It then gradually slows down the car until hands-on is detected again.So the cars have the ability to verify that the operators hands are on the wheel. And the feature requires you to acknowledge that you will keep your hands on the wheel. Yet Tesla doesn't keep a running check that you are keeping your hands on the wheel instead only checking at unspecified intervals. Why enable behaviour (with additional programming to boot (the timer)) that enables incorrect use of your system? Maybe because youtube video showing drivers with their hands on the wheel won't garner millions of hits?
posted by Mitheral at 5:01 PM on June 30, 2016
How long will it take enough people to drive enough distance to get those millions of miles?
It's already done.
There's a solo dev who came up with a system that works reliably in good weather, and he was the only driver contributing to the model, and he was up and running in about a year. He had a car that could keep to its lane on a highway after observing two and a half hour's worth of driving, even around tricky curves.
Google/Alphabet's been working on this around a decade, and they have a fleet of vehicles in constant motion (Google Maps Streetview) and a literal army of coders and entire datacenters devoted to crunching the numbers.
You know that seat-of-the-pants feeling when the car is juuuust about to let loose going around the bend on gravel? That's fluid sloshing around your inner-ear matched with vibrations you feel from the pedals and seat and the way the wheel moves matched to your experience. The self-driver's accelerometer is much more sensitive and precise than yours, as are the sensors everywhere along the suspension, braking and steering systems, and it has the experience of a few hundred drivers, some of them, weighted accordingly in the model, bad-weather rally experts.
The writing in the linked article is breathless and dumb and hero-worshippy in the way early issues of Wired were breathless and dumb and hero-worshippy, but it also lays out how the software works and why it works.
This is just one guy. Volvo is immense. Mercedes is immense. Ford is immense. Google is immense. Telsa has aspirations to immensitude.
I mean, really, once the Lincolns learned to parallel park, it was all over.
posted by Slap*Happy at 7:28 PM on June 30, 2016
It's already done.
There's a solo dev who came up with a system that works reliably in good weather, and he was the only driver contributing to the model, and he was up and running in about a year. He had a car that could keep to its lane on a highway after observing two and a half hour's worth of driving, even around tricky curves.
Google/Alphabet's been working on this around a decade, and they have a fleet of vehicles in constant motion (Google Maps Streetview) and a literal army of coders and entire datacenters devoted to crunching the numbers.
You know that seat-of-the-pants feeling when the car is juuuust about to let loose going around the bend on gravel? That's fluid sloshing around your inner-ear matched with vibrations you feel from the pedals and seat and the way the wheel moves matched to your experience. The self-driver's accelerometer is much more sensitive and precise than yours, as are the sensors everywhere along the suspension, braking and steering systems, and it has the experience of a few hundred drivers, some of them, weighted accordingly in the model, bad-weather rally experts.
The writing in the linked article is breathless and dumb and hero-worshippy in the way early issues of Wired were breathless and dumb and hero-worshippy, but it also lays out how the software works and why it works.
This is just one guy. Volvo is immense. Mercedes is immense. Ford is immense. Google is immense. Telsa has aspirations to immensitude.
I mean, really, once the Lincolns learned to parallel park, it was all over.
posted by Slap*Happy at 7:28 PM on June 30, 2016
« Older Get goin'! *slap* *click* *slap* *slap* *click*... | There's just so much it can't do Newer »
This thread has been archived and is closed to new comments
I've been getting more pessimistic about the time needed to rollout of self-driving cars as I've learned more. Partly based on the technology but these recommendations play into it--they really seem like self-driving would be mandatory to reap the benefits.
posted by mark k at 10:21 PM on June 27, 2016