It's not well-known in the US that we're currently terrible at road safety, so it's interesting to read that Cruise might want to test in other countries. Denmark, Norway, and Sweden manage about 2 road fatalities per 100k people, versus about 13 in the US ( https://read.oecd-ilibrary.org/transport/road-safety-annual-report-2022_badaa1a4-en#page23 ), so I assume "not there". The median among OECD nations in that report is a little below 5.
The human overreaction to the risk of anything new is bonkers. Any online conversation about bicycle safety, someone will eventually start yammering about risk to pedestrians, which is hard to measure because there's no nationwide count of such crash fatalities, but it appears to be about 1 per year, usually in NYC, sometimes SF, sometimes Chicago or DC (2022 and 2023 are outliers, both with I think 3 crash deaths in NYC). The risk to pedestrians is at least an order of magnitude lower than the risk from cars and trucks.
One thing that's interesting to me is the apparent decoupling (in plans, marketing, and regulations) of speed and safety. The pedestrian speed-vs-fatality-risk curves show a steep rise between 20 and 30mph, enough that every 1mph increment in that range clearly matters. This pattern seems to hold for bicycle-pedestrian safety; the recent-ish crashes that I recall usually involve fit and/or hill-assisted young men on bikes, often using Strava to record, share, and compare their speed, probably traveling over 20mph. One or more of the recent NY bike-ped crashes involved an e-bike, so, cue the pearls and fainting couches, get the smelling salts, the new thing killed someone. Snark aside, the bicycle-pedestrian crash death rate will probably rise as more e-bikes hit the road, because the 20mph US assist limit for type 1 & 2 e-bikes is above the normal speed for urban cycling, and approximately nobody on a bike can cruise at the 28mph assist limit of type 3 e-bikes (which do come with more use restrictions). For self-driving cars, size, mass, and speed limits might make a difference, and I am curious about the possibility that the first self-driving long-haul trucks might elect to cruise at a speed near the legal interstate minimum, because that will be much safer than the legal maximum, and for any trip exceeding the distance of a human driving shift, still faster. It would change the traffic flow on interstates, pretty sure the default human driver would be grumpy.
One other safety problem for the self-driving car companies is the understandable focus on legal fault versus avoiding crashes. Suppose the default judgement in any motor-vehicle/pedestrian crash were that the vehicle is at fault; that is, the rule is "don't hit pedestrians, period, no excuses". Thinking this rule through and thinking about what's required to make it happen changes how you bike or drive. There's a Tesla demo video from years ago where they *don't* follow this rule, and the car proceeds w/o slowing or swerving past a man walking two dogs right at the edge of the road. The video runs at 2x, I caught this hazard (with eyes trained on the no-excuses rules) on the first viewing at that speed. If I can do it, they can do it, but the Tesla (or perhaps its driver, not sure if the video was a sham) chose not to, presumably because they would not have been at fault if the man had stepped or fallen into the road. Roadside parking and loading zones cause a lot of problems for this safety change, since those can easily hide people beginning to cross the road. This in turn favors skinnier vehicles, since they can leave more clear space at the edge of the road without crossing the centerline, and favors slower vehicles, since they can stop sooner and if that fails, the crash is still at a lower speed.
You will get what you incentivize (or disincentivize) so it's hard to imagine the companies not having a focus on what the law says creates liability and fault for them. They do understand the court of public opinion is very important, so they won't just try to follow the law, but that'show things will bend.
The answer is, change the law. You could need assign some blame for any pedestrian impact, though I think you would need to add "except extraordinary situations where the pedestrian was clearly the cause" but it's not clear you want to. You can't have these vehicles drive completely paranoid all the time. Such a rule would make them slow down if a pedestrian happened to step towards the street. It's not unreasonable for pedestrians to have some duties too, you just want to set the right balance.
Or you could put a lower amount of liability for hitting pedestrians who act badly, rather than zero. But in the end you need a regime where people can be confident they won't be punished for things beyond their control, because if they can, you kill the whole idea.
Comments
David Chase
Mon, 2024-01-01 09:15
Permalink
Self-driving car safety
It's not well-known in the US that we're currently terrible at road safety, so it's interesting to read that Cruise might want to test in other countries. Denmark, Norway, and Sweden manage about 2 road fatalities per 100k people, versus about 13 in the US ( https://read.oecd-ilibrary.org/transport/road-safety-annual-report-2022_badaa1a4-en#page23 ), so I assume "not there". The median among OECD nations in that report is a little below 5.
The human overreaction to the risk of anything new is bonkers. Any online conversation about bicycle safety, someone will eventually start yammering about risk to pedestrians, which is hard to measure because there's no nationwide count of such crash fatalities, but it appears to be about 1 per year, usually in NYC, sometimes SF, sometimes Chicago or DC (2022 and 2023 are outliers, both with I think 3 crash deaths in NYC). The risk to pedestrians is at least an order of magnitude lower than the risk from cars and trucks.
One thing that's interesting to me is the apparent decoupling (in plans, marketing, and regulations) of speed and safety. The pedestrian speed-vs-fatality-risk curves show a steep rise between 20 and 30mph, enough that every 1mph increment in that range clearly matters. This pattern seems to hold for bicycle-pedestrian safety; the recent-ish crashes that I recall usually involve fit and/or hill-assisted young men on bikes, often using Strava to record, share, and compare their speed, probably traveling over 20mph. One or more of the recent NY bike-ped crashes involved an e-bike, so, cue the pearls and fainting couches, get the smelling salts, the new thing killed someone. Snark aside, the bicycle-pedestrian crash death rate will probably rise as more e-bikes hit the road, because the 20mph US assist limit for type 1 & 2 e-bikes is above the normal speed for urban cycling, and approximately nobody on a bike can cruise at the 28mph assist limit of type 3 e-bikes (which do come with more use restrictions). For self-driving cars, size, mass, and speed limits might make a difference, and I am curious about the possibility that the first self-driving long-haul trucks might elect to cruise at a speed near the legal interstate minimum, because that will be much safer than the legal maximum, and for any trip exceeding the distance of a human driving shift, still faster. It would change the traffic flow on interstates, pretty sure the default human driver would be grumpy.
One other safety problem for the self-driving car companies is the understandable focus on legal fault versus avoiding crashes. Suppose the default judgement in any motor-vehicle/pedestrian crash were that the vehicle is at fault; that is, the rule is "don't hit pedestrians, period, no excuses". Thinking this rule through and thinking about what's required to make it happen changes how you bike or drive. There's a Tesla demo video from years ago where they *don't* follow this rule, and the car proceeds w/o slowing or swerving past a man walking two dogs right at the edge of the road. The video runs at 2x, I caught this hazard (with eyes trained on the no-excuses rules) on the first viewing at that speed. If I can do it, they can do it, but the Tesla (or perhaps its driver, not sure if the video was a sham) chose not to, presumably because they would not have been at fault if the man had stepped or fallen into the road. Roadside parking and loading zones cause a lot of problems for this safety change, since those can easily hide people beginning to cross the road. This in turn favors skinnier vehicles, since they can leave more clear space at the edge of the road without crossing the centerline, and favors slower vehicles, since they can stop sooner and if that fails, the crash is still at a lower speed.
brad
Tue, 2024-01-02 17:19
Permalink
What you incentivize
You will get what you incentivize (or disincentivize) so it's hard to imagine the companies not having a focus on what the law says creates liability and fault for them. They do understand the court of public opinion is very important, so they won't just try to follow the law, but that'show things will bend.
The answer is, change the law. You could need assign some blame for any pedestrian impact, though I think you would need to add "except extraordinary situations where the pedestrian was clearly the cause" but it's not clear you want to. You can't have these vehicles drive completely paranoid all the time. Such a rule would make them slow down if a pedestrian happened to step towards the street. It's not unreasonable for pedestrians to have some duties too, you just want to set the right balance.
Or you could put a lower amount of liability for hitting pedestrians who act badly, rather than zero. But in the end you need a regime where people can be confident they won't be punished for things beyond their control, because if they can, you kill the whole idea.
Add new comment