Separate names with a comma.
Discussion in 'Autonomous Vehicles' started by jkliu47, Jul 26, 2016.
NTSB Issues Preliminary Report for Williston, Florida, Highway Crash
nothing new there
Actually I think there is... Right off the bat: "The report states that according to system performance data downloaded from the car, the indicated vehicle speed was 74 mph just prior to impact, and the posted speed limit was 65 mph."
There are also some good pictures of the impact on the trailer and of the car after the collision.
Actually there is something new. The speed is confirmed at 74 mph.
Previously people either assumed the 65mph as the limit was (and the police report wrote) or above 85mph (as an witness reported the car passed by her when she was travelling 85mph). 74mph is pretty much right in between those speeds.
Also, the detailed report says the vehicle's HV battery disconnected at point of impact, which meant the car coasted into the pole. An earlier eyewitness claimed the car continued to maintain its speed after the collision as if Autopilot didn't disengage.
Good catch. That is important too since it shows the safety systems did detect the crash and autopilot definitely wasn't engaged anymore after the initial impact.
From the description, the initial impact was not that severe (energy wise), which is why the car was able to coast so far.
I would be curious as to how many accidents occur while cruise control, on any car, is engaged. Seems to me, blaming so-called "Auto Pilot" is inaccurate. Cruise control + distracted driver + exceeding speed limit + truck driver error ought to pretty much sum it up. Auto Pilot is being villified incorrectly, IMHO.
This also explains why people might have incorrectly thought that. If the HV was disconnect and the car was coasting, then regenerative braking wasn't in effect. Without regen and someone/something actively pressing the brake, the car is quite efficient and can coast for an extended period of time on flat roads.
With the speed, we can now calculate the time from the crest of the hill to the point of impact.
That is, if I remembered the distance, was it 1/4 mile?
Horrible situation. This is similar to any airline accident where there's so many safety systems and redundancies, there has to be a series of events in an exact order to have a fatality. Here, excessive speed, plus the truck turning, plus the driver not being alert, plus the truck height shaving off the top of the car, the sun, the color of the side of the truck, etc.
Any of those factors not present could've meant an accident, but not a fatality.
Sorry for the family's loss.
Fortunately, the NTSB isn't the mindless media. They have a long history of properly investigating aviation disasters where human-automation coexistence is frequently an issue.
I actually look forward to what they have to say. It sounds like they are focusing on why AEB didn't stop the car (and whether or not that's acceptable for the automotive industry in general), but I also hope for their to be a human behavior narrative too on whether or not humans are inherently bad at supervising a system like Autopilot that can appear to do the right thing for hours at a time.
Maybe it's difficult to discern, but if AEB had applied braking...what speed would it have reduced the Model S before impact?
Seems to me, looking at the damage to the trailer, that even moderate speed (~35mph) would still have made for a pretty bad day.
I thought I read somewhere (perhaps in the Manual's PDF) that AEB will only reduce speed by a maximum of 20 mph.
I remember similar, just wondering at what initial speed 20 mph is achievable. Seems like 74 mph may be a bit high to achieve it.
Updated Owner's Manual - Interesting Information | Tesla Motors
Sounds like the Model S would have hit the truck at 49mph, had AEB engaged. At 49 mph, I don't think AEB would have changed the outcome. And based on where it went under the truck, the half a second of distance the trailer would have moved laterally wouldn't have changed the outcome either.
So, short of the driver seeing the truck...it doesn't seem the result could have been any different: even in the best case scenario.
Except that at the lower speed, the truck would have moved further forward, and he might have hit the rear axles instead of being decapitated. That would be totally survivable (in a Tesla...).
He hit nearly dead center, and the semi wasn't moving very fast...maybe someone can calculate it...but AEB would have to engage pretty early to make a difference.
That's more of a lottery-winning argument. Maybe if he drove at 120mph he would've hit the cab and survived too …
AEB probably would reduce the severity but it has to respond relatively early to the act of the truck turning onto the road and a realization it wouldn't complete the turn in time. That's a hard engineering problem to solve.
I suppose the AEB might have engaged the driver enough to duck (+ given him extra time to do so). By engaged the driver, I mean call his attention to the situation at hand.
AEB would never be calculating the truck's speed and figuring if it would make it in time. Most appropriately, the car would have slowed as soon as the truck crossed the forward path of the Tesla. Even 100 feet away. Then increasing braking until full on slam when collision was imminent. I guess technically this would be autopilot, but a gradual slowing would be really nice to call attention to the driver. If the decision was incorrect - ie the truck moving quickly enough, the driver could override by hitting the accelerator. That would be ideal. Just like we do when autopilot slows for an exiting car and we are ready to move past them.
The scenario is an edge case. Would be nice to see side skirts as the end result. With any sort of diesel/carbon tax, the side skirts are a no brainer. Safety benefit to boot. Would also like to see mandatory sensor change that can "see" higher.
Did he go under the trailer? between the rear axles of the Semi and the axles of the Trailer? Like Fast'n'Furious style but laterally not perpendicular? Would those Trailer Aero-Guards have helped the radar pick up a signature? Know what I'm talking about? those flat sheets to increase efficiency for wind?