The footage from the normal camera does matter in that it's the main way that we (humans) can process the scene. The parent comments are just pointing out that the camera footage is likely darker than the actual scene in person.
Waymo cars are capable of sensing vehicles and pedestrians at least half a block away in every directions. I was reserving any judgement on wether this collision could have been prevented, but seeing the video tells me that 1) a human driver might have hit the victim regardless, and 2) I'm very surprised that the LIDAR sensor didn't cause the car to stop to a halt much, much earlier. This is exactly the kind of situation that I would expect self-driving cars to be better than human drivers.
I agree that dashcam/external cam footage is going to be limited and possibly misleading, and I would think/hope such footage isn't the primary factor in evaluating accident cases. But I do think there's value to it. I shouldn't have said that it is the "main" way for us to process a scene, but the most accessible/relatable way.
What you posted looks pretty cool, I don't know enough about it to understand what I should be prioritizing focus on, but we can chalk that up to ignorance. The benefit that driver-view footage has is that it is a viewpoint all of us are familiar with. If you ask me to watch dashcam footage to assess some kind of traffic thing, there's a general expectation of where I keep my eyes and what I notice.
This normal-human-view mode is probably going to be necessary in AV cases in which we determine whether the car's AI did the right thing. Presumably, as AV becomes mainstream and extremely safe, these accidents will involve edge cases and outliers which are poorly interpreted by sensors/non-human-vision. Seeing the scene as a human driver does might be a necessary starting place?
But the Uber case in AZ, IMO, proves your point. The Tempe police quickly made a judgement call based on what seems to be inadequate video. Everyone who can now view the video will also be inclined to think how impossible it would be to avoid hitting the victim, even if the actual scene in-person has much more light. And of course, we don't want to judge AV solely on whether it performs as well as normal humans.
Uber may not be at fault, legally speaking. That's up to the legal authorities to decide.
However, as a society and civilization, and even more so, as engineers and scientists, we are going to expect that the autonomous car matches or exceeds human-level performance in critical situations like this.
Therefore the time spent on investigating, understanding, and discussing the root causes of the accident is worth understanding. Accidents like these generally do not happen due to a single factor. It is necessary to understand all the necessary factors if we want to make autonomous driving systems more reliable.
At the very least we need to understand whether the pedestrian appeared in the other sensors that a human could have identified by looking at the sensor data, and if yes, whether the autonomous system matched or exceeded human-level performance by detecting the pedestrian, and if the pedestrian was indeed detected, why the autonomous driving system failed to respond to the situation.
Surely not? Cars are routinely driven by people who are not owners, and liability for traffic offences (including that the vehicle must be insured) is with the driver.
In my experience typically only minor infractions like parking violations are assigned to the registered owner of the vehicle, but in other case – accidents, running red lights etc. – the driver is liable regardless of who owns the car.
No. The general rule is that negligence is required to be held responsible. If I let my next door neighbor borrow my car to go to the grocery store, and he hits someone, I'm not responsible. Unless, the person can prove "negligent entrustment", i.e. it was irresponsible just to let this person borrow my car, e.g. they're a habitual drunk, or blind, or 11.
However, most auto liability insurance covers whoever you permit to drive the vehicle, so the owners policy does typically cover the fender bender on the way to the grocery store.
Correct, the owner's insurance policy is the primary coverage when the owner lends their car to a 3rd party. Obviously in the case of a moving violation the driver is at fault and receives the penalty, but damage is still covered by the owner's policy. In the case where the other driver is at fault, that car's owner's insurance is liable.
exactly this. what's the response time of software? it ought to be close to zero and significantly faster than human's. let's say it's a generous 0.5s - no brakes where applied at all, and even with the crappy darkened video we got (place isn't that dark https://www.youtube.com/watch?v=1XOVxSCG8u0 ) the pedestrian was in view for 2 to 3 seconds.
car didn't see it at all even in those last moments.
Well it was a pedestrian but they were walking their bike across the road. It's not like the software should make a distinction between a cyclist in the way and a bicycle with no rider in the way.
Indeed, it's hard to find pedals without them. Even ones that cost $10 a pair have reflectors. Unfortunately, pedal reflectors are ineffective when the bicycle's path of travel is perpendicular to the light source. The video doesn't reveal evidence of other reflectors, such as the common spoke-mounted ones whose purpose it is to highlight a bicycle traveling crosswise. For a moment, the bicycle is clearly illuminated by the headlights; I don't see any spots of light on the wheels or elsewhere.
For a side view, the reflectors on the tires (visible at the end of the video) are way better indicators of “watch out! Bicycle” than those reflectors.
See this video for a comparison of visibility (not in English, but that's immaterial - set speed to 2x ;)): starting with a "bike ninja" and going all the way to "Christmas tree" https://youtu.be/oAFQ2pAnMFA?t=1m0s
It's from 2011, there's been a lot of improvement in consumer-grade cameras since. Even so, it fits my perception IRL: even a small reflector is orders of magnitude better than no reflector, and adding multiple (esp. covering 360 viewing angles) makes you stand out at night; same goes for pedestrians.
This 100%. When I drive, I watch the road. I don't watch my mobile phone, I don't watch the kids behind, I don't watch my wife. I don't watch the sky. I don't watch the GPS.
I just watch the road in front of me.
My idea is that the car has been behaving well for a long time and consequently the driver lowered is vigilance. Big mistake.
A fully attentive human driver might have hit this person regardless. Would they have hit them while taking no evasive action whatsoever? No swerving, no brakes?
I don't think so: the dash cam video is misleading. I had multiple ninjas jump at me before, and although I did notice and avoid them, they were not visible on the dashcam until the very last moment. Surely Uber would not release data to intentionally mislead the public?
Even so, I count a full second from when a human paying attention would have seen something just using this video as eyes, until impact. The stopping distance at 35mph is 136ft, which is 2.65 seconds at 35mph, so the accident would still happen but the impact speed could be lower.
Yeah, but at that speed, it's more than possible to swerve around an obstacle rather than screeching to a halt before touching it. Even turning slightly to the left/right would have made a dramatic difference in the outcome to this person's life. Not to mention the person in the car that might have also been severely injured if this was a heavier obstacle.
This was purely bad software, and no failure scenario being programmed in. I really don't think it's that difficult to program split-second reaction to obstacles that appear into the driving path. We need to get to a point where these vehicles can do stuff like this, even in a 2-dimensional way:
They seem to use 2.5 seconds as the standard for drivers to perceive and react to an obstacle, which based upon studies covers 90% of all drivers. 1.5 seconds to perceive, 1 second to react. Then you have maneuver time on top of that 2.5 seconds.
Given this, 1 second seems very low. A large percentage of drivers would probably plow into them at full speed.