Surprise: Automated driving biz finds automated driving safer than letting you get behind the wheel
- Reference: 1615361343
- News link: https://www.theregister.co.uk/2021/03/10/waymo_automated_driving/
- Source link:
Chandler, a popular venue for testing automated cars in the United States due to its favorable weather and minimal regulation, became a Waymo test ground in 2016. It's about 16 miles from Tempe, where in 2018 an Uber self-driving car collided with and killed pedestrian Elaine Herzberg, an incident for which safety driver Rafaela Vasquez was [1]charged last year with negligent homicide and has pleaded not guilty.
[2]
Trent Victor, director of safety research and best practices at Waymo, said in a [3]blog post on Monday that the company's latest research supports data released in October, 2020, that showed its AI system, referred to as the Waymo Driver, was only involved in minor collisions after six million miles on the road.
The automatic auto firm recreated 72 fatal crashes in 91 simulations, putting the Waymo Driver in the role of both the initiating driver and the responding driver for accidents involving two vehicles. And in the simulator at least, the company's software outperformed the humans involved in the recreated incidents.
[4]
"When we swapped in the Waymo Driver as the simulated initiator (52 simulations), it avoided every crash by consistent, competent driving, and obeying the rules of the road—yielding appropriately to traffic, executing proper gap selection, and observing traffic signals," said Victor.
1/3 Today, we release results that show how the Waymo Driver likely would have performed in the majority of fatal crashes that occurred on the same roads over a 10 year period. This builds upon the research we released in October 2020. Read our findings: [5]https://t.co/RrIdwfWoBb [6]pic.twitter.com/CtGsiYgqnY — Waymo (@Waymo) [7]March 8, 2021
The implication is that humans could do as well, if they actually obeyed road rules, made a concerted effort to drive carefully when behind the wheel, and were never distracted or impaired.
When the AI system acted as the responding driver, it managed to avoid a collision 82 per cent of the time and in an additional 10 per cent of incidents, it took action that reduced the severity of the recreated crash.
In the eight per cent of simulations as a responding driver that Waymo's code couldn't improve on, each was the result of a rear-end collision – which human drivers also have a hard time avoiding.
Machine learning gets semi conscious... Waymo, Daimler vow to bring self-driving trucks to American highways [8]READ MORE
Victor argues that because 94 per cent of crashes involve human error, based on [9]NHTSA data , Waymo's robotaxis have the opportunity to improve road safety.
The research paper
[10]PDF
describing Waymo's findings is a bit more circumspect, allowing that the tests have limitations and real world outcomes may differ. For example, the paper notes that the simulations did not include any other vehicle traffic, so any effect such traffic might have on the Waymo Driver's sensors is not modeled.It also observes that "the collection of potential failure modes for [automated driving systems] may be different than that of a human," which alludes to potential problems arising from sensor or electronics snafus that would not hinder a human driver.
There are currently [11]more than 60 companies involved in self-driving tests or having permits to do so in California. As one of those, Waymo is obligated to report its disengagement rate, the rate at which a human driver has to take over from the Waymo Driver.
The number Waymo reported last year was one disengagement per almost 30,000 miles, an improvement from one in 13,219 miles in 2019. The mech chauffeur biz, nonetheless, has [12]argued that disengagement rate is not representative of the capabilities of its system.
[13]
Waymo is still working on [14]driving in snow . ®
Get our [15]Tech Resources
[1] https://www.theregister.com/2020/09/16/self_driving_uber_homicide_charge/
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2YEim1KJfvcCmQxZVQCNugAAAAEE&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://blog.waymo.com/2021/03/replaying-real-life.html
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33YEim1KJfvcCmQxZVQCNugAAAAEE&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://t.co/RrIdwfWoBb
[6] https://t.co/CtGsiYgqnY
[7] https://twitter.com/Waymo/status/1369016501932466178?ref_src=twsrc%5Etfw
[8] https://www.theregister.com/2020/10/28/waymo_daimler_ai/
[9] https://www.nhtsa.gov/technology-innovation/automated-vehicles-safety
[10] https://storage.googleapis.com/waymo-uploads/files/documents/Waymo-Simulated-Driving-Behavior-in-Reconstructed-Collisions.pdf
[11] https://www.dmv.ca.gov/portal/vehicle-industry-services/autonomous-vehicles/disengagement-reports/
[12] https://twitter.com/Waymo/status/1232758355082375168
[13] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YEim1KJfvcCmQxZVQCNugAAAAEE&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[14] https://twitter.com/Waymo/status/994091703379415042?s=19
[15] https://whitepapers.theregister.com/
Re: But... but... we are driving because we like it, right?
I mean, who doesn't enjoy being stuck in traffic for an hour every morning?
Do you think the wage slaves who have to be at work at spot on 8:59 will be able to afford driverless cars?
Re: But... but... we are driving because we like it, right?
Do you think the wage slaves who have to be at work at spot on 8:59 will be able to afford driverless cars?
You pay your mill workers enough that they can own a car?
Times change.
How many of the simulations involved a pedestrian walking across the road?
You mean a carrier bag? Or was it a rock?
Dept of the Blindingly Obvious
Of course there's fewer incidents if everyone obeys the traffic rules.
It's like saying if everybody obeys the 'Murder is not nice, don't do it' rule then fewer people are killed......
But, People as a collective group are stupid, just look at the Darwin Awards.
Different defects
Human driving depends on the fiction that people can continuously pay attention to the road ahead (which itself is known to be false) while at the same time taking note of traffic and direction signs, the rear view mirror, the speedometer and so on. Learning to drive pushes a lot of this impossible workload into the subconscious mind. But when our own automated natural intelligence goes wrong, it can result in accidents where people apparently haven't seen a pedestrian or cyclists that was in plain sign, and they have no idea why.
AI doesn't suffer from this problem, and neither should it deliberately disobey traffic rules as most human drivers do from time to time. but it has two other problems. Firstly current implementations seem to make more mistakes than humans in categorising what is in vision. This will presumably improve. But secondly, not actually being intelligent, it cannot grasp what is going on in a situation, in terms of the intentions and likely actions of human participants. This would be a big step beyond curent AI. I suspect that in complex situations such as inner city driving we will need to rely on human intelligence for a long time yet, and this is where the accidents involving pedestrians and cyclists occur. For these situations some automated assistance to human driving is available, but that brings the problem that humans hate making effort, and the more help you give them, the less work they do themselves. We will need some good psychology to overcome this.
Re: Different defects
There is nothing like driving a rhd car in a lhd country to see just how much of the driving workload is automated by early training...
Re: Different defects
Know what you mean! I'm a Limey and I have to actually think when I drive in Spain rather than going by instinct. I feel like I'm driving in some weird alternative mirror universe.....
I find that having sat-nav helps me concentrate on the driving, although you have to be careful as there was a stretch of road I had to drive along where you had to do the opposite of what the sat-nav was telling you. E.g. "turn left", nope there's a no entry sign there, I'm going to have to turn right....
Also I'll find myself at roundabouts saying "on does trays"....
Re: Different defects
Been living in a drive on the right country for the last 11 years. I still, very occasionally, find myself momentarily on the wrong side of the road after a right turn.
Re: Different defects
I'm not sure that intelligence is needed to predict the motion of most groups - particularly when you can substitute an enormous set of observations.
People, whether in cars, on foot, on motorbikes, pedal cycles, scooters, skates, are basically predictable - often due to simple physics, a vehicle moving cannot suddenly translate laterally, it must adjust it's angle and do so. Similarly people can't instantly disappear, much to the chagrin of many people who currently control heavy machinery on the roads.
Pedestrians are probably the least predictable group, since they can stop to concentrate on something unrelated.
False comparison and policy based evidence
" its simulated AI driver isn't a careless, distracted jerk ".
Nor in fact are most human drivers. The actual incidence of accidents per vehicle mile is very low, even including minor prangs. I'm also (with quite a bit of experience of them) very wary of simulations (particularly simulations by interested parties). Furthermore, until (unless) full autonomy is achieved, the "safety net" of a human driver taking over in emergency is utterly illusory. Reaction time from total inattention (as of an uninvolved passenger) is substantially longer than from partial attention (as of an inattentive but involved driver).
The accident simulation's conclusion doesn't make sense. Most human drivers would also pass the test. It's the integration over time that counts.
But... but... we are driving because we like it, right?
I mean, who doesn't enjoy being stuck in traffic for an hour every morning?