I completely disagree with this. The pedestrian squared his/her body, and made eye contact sufficient with the near lane to come to the stop - how does the driver not see the other driver stopped at the crosswalk?
I dunno, I’ve been watching and rewatching trying to decide if I would have been able to stop before the crosswalk in this situation. The oncoming car is blocking any view of the pedestrian for the first two seconds, and the vehicle in question is in the crosswalk in about 4 seconds. That’s a pretty small window to respond. I think the car should have slammed on the brakes as soon as it detected somebody crossing, but it’s a close enough call that I’m not entirely sure the car didn’t make the right call.
Eh…well…our town has those crosswalks all over the place, and I tend to be a careful driver, so maybe it’s just habituation.
Around here it is normal ped behavior (and I do a LOT of pedestrianing myself) to step off the curb timed to pass behind an on-coming car in the far lane(s). Having them then stomp their brakes creates an unsafe situation for everybody.
As to the Tesla vid, when the ped steps off the curb “our” car is already into the time/place where only max effort braking will stop short of the crosswalk. That would be an appropriate reaction for a an immovable object, a clueless child or slow-moivg elderly. For a normal adult, nope; they’re maneuverable enough on their own and they’re safer not having a skidding mostly out of control car near by. Although if I was manually driving the Tesla I’d have my eyes glued on her to ensure she focused in my direction before she got much farther.
In all what’s going on is not absolutely safe. But IMO it’s the most safe of a shortlist of less-than-fully-safe choices.
I probably would have stopped, because in my estimation the guy would almost be to the middle by the time I arrived. That would mean I would have overrode FSD and forced a stop. It’s hard to judge distance, but from 28 MPH, I think the car could have stopped before entering the cross walk.
Me, too. I cross several sets of mid-block cross walks everyday on my commute to work, and most people don’t stop unless they would actually hit the pedestrian. I stop (give me a cookie!).
The one that much more frequently has pedestrians is at a T-intersection to the main road. It is particularly dangerous for pedestrians. Cars will stop on the main road to let a pedestrian cross. Cars on the side street will think the cars have stopped to let them in, and will then turn right across the cross walk, ignoring the pedestrian. Point of the story is the legitimate what-aboutism of human drivers still being bad, even after 100+ years of development.
FSD has to make predictions about pedestrians, and it doesn’t seem to be too good yet. So far, I’ve only noticed errors where FSD slows or stops because it interprets the pedestrian’s (or bicycle’s) behavior as being about to enter the street, but I don’t interpret it that way. For example, a bicyclist starting to move and initially facing to cross in front of me (FSD stops), but the bicyclist is looking and initiating a turn to cross parallel to me (no stop necessary). Along with the body language, added context FSD ignores is that by turning the bicycle is crossing with the light, and not against the light directly into moving traffic.
One of my gripes with FSD at the moment is I don’t think it is quite aggressive enough when stopping for yellow lights. A few times recently I’ve had to stop hard because FSD was clearly going to just go for it, and enter the intersection on red. Problem is, it could have been normal stop for a yellow from a distance, but waiting a bit to see if FSD will stop means that by the time I decide to stop, it is has to be a much harder stop.
I think the yellow light behavior can be much more deterministic. The car knows how far to the light, how fast it is going, and how much time it has to stop or go through the intersection. Pick something. If it can’t clear the intersection in 4 seconds (or whatever), stop.
Yes, it can be…
The duration of any particular yellow light is unknowable. Absent that variable it’s impossible to calculate where the vehicle will be at the yellow/red transition if it attempts to go, not stop, when the light turns from green to yellow. Humans face the same problem.
The upside for a computer is that presumably it will notice the green/yellow transition at the instant it occurs, unlike a human which might not pick up on that for multiple seconds. Not realizing a yellow is “stale” should be a human-only problem. And it’s a problem that now results in many honest red light runs (as opposed to deliberate, dishonest red light runs).
The NHTSA standards say that yellow lights should be long enough that a car traveling at/near the speed limit should be able to either stop normally or proceed at a constant legal speed and either stop short, or enter the intersection before the yellow ends. Which leads to the decision-making rubric most of us humans use. Namely:
If we see the light transition to yellow, and much above normal braking would be required, then proceed at speed and the yellow should be long enough you’ll be legal and safe. If much above normal braking is not required, then stopping is the necessary choice.
The only individual variation is our personal assessment of “much above normal”. Which obviously can be defined as a rigid software parameter that’s consistent across the self-driving fleet.
Except for the jurisdictions that have decreased yellow light timing in order to improve red light camera revenue. (I hope that isn’t happening any more.)
I think there can be reasonable boundary conditions. A yellow light is unlikely to be longer than 6 seconds, so if continuing means clearing the intersection in 5.5 seconds, then the car should probably stop. If the car will be across the intersection in 2 seconds, then definitely go.
As for knowing, I think this is where self driving cars could have real advantages. Anytime a car encounters a yellow light, record the length of time it is yellow (if conditions allow), and then feed that back to be integrated into the map. Even a data point of “at least 3.8 seconds” is more useful than nothing.
Better yet, get cities on board with just publishing yellow light timing data.
And even better still, Rwanda style traffic lights with integrated counters to show how long until they change.
I don’t think it’s just you. I had the same reaction…the driver should have seen the pedestrian coming and should have stopped. I’m not happy that the car kept going.
Yellow time should have nothing to do with it. The rubric for a yellow light is, as @LSLGuy said, “stop if I reasonably can before the line”. That’s it.
I’ve heard for much of my driving life that if a pedestrian is anywhere within the crosswalk area, the driver must stop. And the linked article implies as much. But reading the actual law, that doesn’t seem to be the case:
(a) The driver of a vehicle shall yield the right-of-way to a pedestrian crossing the roadway within any marked crosswalk or within any unmarked crosswalk at an intersection, except as otherwise provided in this chapter.
(b) This section does not relieve a pedestrian from the duty of using due care for his or her safety. No pedestrian may suddenly leave a curb or other place of safety and walk or run into the path of a vehicle that is so close as to constitute an immediate hazard. No pedestrian may unnecessarily stop or delay traffic while in a marked or unmarked crosswalk.
(c) The driver of a vehicle approaching a pedestrian within any marked or unmarked crosswalk shall exercise all due care and shall reduce the speed of the vehicle or take any other action relating to the operation of the vehicle as necessary to safeguard the safety of the pedestrian.
(d) Subdivision (b) does not relieve a driver of a vehicle from the duty of exercising due care for the safety of any pedestrian within any marked crosswalk or within any unmarked crosswalk at an intersection.
The pedestrian has right-of-way–but that does not imply that all cars must stop, even if the pedestrian is far away. Unprotected left turns must also yield right of way to other cars, but that just means that a car is allowed to make a turn if it can do so safely and before any oncoming traffic arrives–it doesn’t mean that turns are illegal if there’s oncoming traffic a mile away. Likewise, I see nothing here that implies a car must stop if it’s apparent that passing through will not impede the pedestrian’s route. The pedestrian was moving at a consistent speed, and the car slowed and exercised its version of due caution (a human might have been hovering over the brake, but the computer doesn’t need to).
At any rate, I feel better now about making right-turns-on-red when there’s a pedestrian way on the other side of the intersection. Or when going past a crosswalk after the pedestrian has crossed my path, but is still within the intersection.
The colloquial version I was taught in driver’s ed was:
If you make the ped break stride, you screwed up for sure. If you expected them to break stride but they didn’t you also screwed up.
This was in California, one of the bastions of “Peds come first”.
Heh, I hadn’t heard that. By the time I took driver’s ed, I think they might have been teaching kids to be a tad less aggressive. In fact, they pretty much drove in the message that if you hit a pedestrian under any circumstances, crosswalk or not, it’s your fault.
I’m not sure that was a good approach, since obviously it’s impossible to drive as if a pedestrian might leap out from behind a bush at any moment. But if everything is treated equally, that might cause people to increase risky behavior at crosswalks, as opposed to decreasing it elsewhere.
Recognizing that the “break stride” rubric had the advantage of making different allowances for brave teen boy peds vs little old ladies with canes or Moms herding multiple 6yos.
The goal wasn’t to brush them all back, but to ensure you as driver were sufficiently courteous to them that you didn’t scare/startle them. At the same time, peds who jumped out with cars nearby tended to be rewarded with skidding cars (usually) barely missing them. In the military we called that “self-critiquing behavior”.
It seemed to work well on both sides.
Yeah, it’s a decent rule of thumb that matches the law. Pedestrians aren’t allowed to act like assholes, either (see section (b) above). But they are allowed some variance in walking speed, etc., and the driver should account for that.
One of my college friends and classmates used to freak me out when we walked together by striding boldly out into roadways without really looking or verifying the oncoming cars saw him. When I expressed dismay he said dismissively, “Get across or get money”.
I’m not worried so much about hitting the pedestrian as I am about the tailgating dickhead 3 feet from my ass who likely doesn’t know, or care, about the crosswalk laws (and whom I cannot assume sees said pedestrian). Do self-driving cars take such things into account, I wonder?
I have no idea about self-driving cars. But every slowdown I make on surface streets or freeways is made with half my attention on the problem ahead and the other half on the problem behind. Often I’m prioritizing avoiding being hit from behind more than I am avoiding the problem ahead.
I’ve seen no evidence that Tesla’s system does this in any direct way, but I’m not sure how I could tell the difference.
What is a factor is that the better “reflexes” can mean the car starts slowing down earlier for a stop, which means a gentler deceleration, which means you’re less likely to be rear-ended. I was in traffic just recently and Autopilot started slowing before I noticed cars were coming to a stop, quick. It still had to brake pretty aggressively, but not as much as I’d have had to with an extra second of latency.
My wife was read ended twice in Taiwan where drivers are much more agressive. Cars will continue entering an intersection well after the light turns red.
The first time, the light was turning red when a driver coming the other way turned left in front of her. She stopped to avoid a collision but the car behind her was accelerating to beat the light and ran into her.
The second time was different scenario, but similar in that the car behind wasn’t watching her, but being aggressive.
Insurance fixed the car, but we lost value when we had to sell the car when we left Japan.