US Investigates Fatal Crash of Ford EV With Partially Automated Driving System (apnews.com) 41
America's National Transportation Safety Board "is investigating a fatal crash in San Antonio, Texas, involving a Ford electric vehicle that may have been using a partially automated driving system," reports the Associated Press:
The NTSB said that preliminary information shows a Ford Mustang Mach-E SUV equipped with the company's partially automated driving system collided with the rear of a Honda CR-V that was stopped in one of the highway lanes.
Television station KSAT reported that the Mach-E driver told police the Honda was stopped in the middle lane with no lights on before the crash around 9:50 p.m. The 56-year-old driver of the CR-V was killed. "NTSB is investigating this fatal crash due to its continued interest in advanced driver assistance systems and how vehicle operators interact with these technologies," the agency statement said.
Ford's Blue Cruise system allows drivers to take their hands off the steering wheel while it handles steering, braking and acceleration on highways. The company says the system isn't fully autonomous and it monitors drivers to make sure they pay attention to the road. It operates on 97% of controlled access highways in the U.S. and Canada, Ford says.
Television station KSAT reported that the Mach-E driver told police the Honda was stopped in the middle lane with no lights on before the crash around 9:50 p.m. The 56-year-old driver of the CR-V was killed. "NTSB is investigating this fatal crash due to its continued interest in advanced driver assistance systems and how vehicle operators interact with these technologies," the agency statement said.
Ford's Blue Cruise system allows drivers to take their hands off the steering wheel while it handles steering, braking and acceleration on highways. The company says the system isn't fully autonomous and it monitors drivers to make sure they pay attention to the road. It operates on 97% of controlled access highways in the U.S. and Canada, Ford says.
Mach-e question (Score:2)
I haven't been around one of these yet. Do they use the car's speakers to simulate a more throaty engine noise like they do in the later models of gas Mustangs?
Re: Mach-e question (Score:2)
A bit of an electric whine (motor not whine), it has an optional game acceleration noise.
Re: (Score:2)
Re: (Score:2)
If you think that's dumb, my previous ICE car used its computer to manipulate the CVT into shifting harshly on purpose, because the designers believed people didn't want to get used to something different. The "feature" could not be disabled.
It took buying a used Chevy Bolt to finally be truly rid of shift points.
Welp (Score:4, Insightful)
The vehicle that was crashed into "was stopped in the middle lane with no lights on before the crash around 9:50 p.m."
Is any automated system going to be able to handle that? Most humans wouldn't.
Re:Welp (Score:5, Interesting)
Your comments about humans show why we suck, drivers are not supposed to out-drive their headlights... If it is dark slow down.
Re: (Score:2)
And were there cars in the adjoining lanes to allow a lane change? Sometimes the least-bad solution is to hit the thing.
Oh, it's a Ford. Never mind.
Re: (Score:2)
Yes it would be able to see it quite a ways off on a straight road. A number of big trucks/busses have had radar + warning + emergency braking on them for a long time. If it was stopped on a curve (so not exactly in direct front of the oncoming) then that would make radar or cameras job more difficult to realizing it was in your lane of traffic.
Human, or automata? [Re:Welp] (Score:5, Informative)
Yes it would be able to see it quite a ways off on a straight road.
And I-10 is indeed straight near Foster Road in San Antonio, where the accident took place.
Yeah, this would be a completely understandable accident if it were a human driver (car stopped in middle lane of expressway, at night, with no lights on? Yow.) but you'd think an automated driving would be good at detecting this. But do note, the article say the car "may have been using a partially automated driving system." So it's not actually clear that it wasn't a human driving.
Re: (Score:2)
Re: Human, or automata? [Re:Welp] (Score:2)
Re: (Score:2)
-Darn it, click-baited again...
Re: (Score:2)
The front collision prevention system is quite limited though. If the car was doing 70, it wouldn't detect the obstacle or apply enough braking to avoid hitting it. It might have lessened the impact by applying the brakes.
The range is limited to prevent false positives when approaching bends.
If there was a failure of the autopilot, it was that it didn't monitor the driver well enough and they were able to stop paying attention for long enough to not avoid the stopped car.
Human driver (Score:2)
I was on an Interstate when I was startled by a car stopped in, I guess you could call it an island, where highway lanes merge.
I guess it was the correct course of action to keep going rather than to stomp on the brakes, which would have risked a chain reaction of rear-end collisions, or to swerve into a lane to the left, where people were driving with not enough headway to allow this.
I guess I could have been more vigilent with my visual scans. At the time I was talking with my passenger about which
Re: (Score:2)
Or there was a vehicle in front of them which saw the stopped car just in time and swerved, not allowing sufficient time for the driver to take back control. I don't know about the Ford driver assist system, but F-111 fighter pilots have to be trained to keep their hands off the yoke while doing terrain-following under autopilot, because any movement will cause it to disconnect and humans cannot react to changes in terrain fast enough.
Radar not as good as visual, even in the dark. (Score:1)
Radar doesn't care about sunlight. Most car's pre-collision system should have detected it.
At highway speeds, radar systems are not going to be able to detect that car in time top avoid a collision.
As we plainly saw in this case, the Mach-E does have a radar module.
It also depend on what the other car was made of, how quickly it could even determine there was a car there... I think there's a reason why Tesla ditched relying on Radar for self driving, vision is just inherently better and the radar data may
Re: (Score:2)
Radar's biggest downside is that it can't tell the difference between an overhead bridge in the distance as you approach the crest of a small hill, a car completely stopped in the lane in front of you, and a coke can sitting in the road with the concave bottom facing you. All three will look like large object on the road. The usual trick done with radar is to simply ignore anything that is not moving with relation to the background.
This is why Teslas used to run into parked firetrucks and construction veh
Re: Welp (Score:1)
LIDAR and cameras (most likely the latter) still have to have line of sight to detect something. There are tons of variations where nothing, no human, no camera, LIDAR would have detected this, hilly terrain, other obstructions in front of you. Especially given it is unexpected, given there were no hazard lights on, given it was dark. That is why you move your car out of the roadway when it breaks down, even if you have to push it there before you call for a tow.
Re: Welp (Score:2)
Re: (Score:3)
The vehicle that was crashed into "was stopped in the middle lane with no lights on before the crash around 9:50 p.m."
Is any automated system going to be able to handle that? Most humans wouldn't.
Except that should be the very instance the radar systems (which this car has) should excel at. Radar doesn't use sunlight to work like camera systems or human eyes.
Re: (Score:2)
Does it have radar? Ultrasonic parking senors, sure, but usually those (per Bosch: https://www.bosch-mobility.com... [bosch-mobility.com] ) max out at 5.5M or about 18' That isn't much help at freeway speeds
Near as I can tell (from https://en.wikipedia.org/wiki/... [wikipedia.org] and various Ford press releases), the Mach-E has EyeQ4 which is camera only (with 1 or 3 cameras depending on which version it has), so it is very possible that there wasn't enough light to determine that there was an obstacle in the road, especially if the stopped v
Re: (Score:2)
Most humans wouldn't.
It's called having functional headlights and paying attention to the road.
Re: (Score:2)
Well, this is why an investigation is needed.
If you had a system with LIDAR or high definition radar, then it might have been able to avoid the collision. With current typical radar, the car wouldn't have been able to determine 'stopped car' until the cameras caught it...
Now could the cameras have caught on? Yes, the car is stopped (foiling radar since radar has to toss stationary objects unless correlated by some other system due to false positives) and lights are off (foiling passive cameras), but it sho
musk bad (Score:1)
Re: musk bad (Score:2)
But.. we are seeing this plastered... Your comment doesn't make sense and you needed this plastered on the news in order to have a chance to comment on it...
Re: (Score:3)
Anyone willing to bet there was NO RADAR? (Score:2)
Even if some Fords have is as an option since 2007 (at least) it can very well be the case that they don't have it in these much fancier somehow-self-driving systems. Cost saving, of course.
Teslas don't have it anymore in the cheaper models. No ultrasonic parking sensors too. They're doing everything with cameras, not that many too! Sure, there are blind spots but they say they can compensate by just keeping track of everything as the car moves. Bad luck for anything that doesn't keep a fixed position like
Re: (Score:2)
I have always questioned the wisdom of any automated driving system depending solely on cameras. Automated driving needs LIDAR, radar, or something else. The issue is how do you tell the difference between a large white object and fog?
Speaking of fog, I also question the wisdom of any automated system that lacks manual fall-back ...
Is it possible the automated system disengaged in a dangerous situation, and the driver wasn't ready for it? Just speculation at this point.
Re: (Score:2)
What is low resolution doppler radar supposed to do exactly? It has lots of stuff in its beam which isn't moving, the car is nothing special to it. Traditional cruise control radar is not imaging radar. Unless you use a fancy "4D" 60 GHz imaging radar, it won't help.
rating the AI (Score:3)
The point is that eventually, on a per km basis, is AI *better* than a typical human? If so, time to stop driving and be driven instead.
AI will still cause some mind-bogglingly dumb accidents that get all the Luddite hand-wringing headlines. But a human running a schoolbus stop sign and mowing down a couple kids somehow doesn't get anyone to say "If only AI had been in charge, this wouldn't have happened".
It's simply a numbers game as to when that threshold has been crossed and then we should all stop driving. Anyone have any idea where we are on that timeline?
Re: (Score:2)
Level 3+ or steer yourself.
Level 2 is dangerous idiocy, it assumes an inhumane capability to pay attention when all the work of driving is taken away from you.
Re: (Score:1)
AI Must be a Lot Better (Score:2)
The point is that eventually, on a per km basis, is AI *better* than a typical human?
Just requiring an AI to be better than the average human driver is setting the bar too low since that means that at least half the people letting the AI drive would do a better job of it than the AI. I'd set the threshold at the 95-99% level i.e. an AI has to be better than 95-99% of human drivers out there. This is because we are not going to trust an AI system unless we know it is well above a typical human's driving capabilities. We all tend to think of ourselves as good drivers and anything less than a
New meaning (Score:2)
This gives a new meaning to Found On Road Dead.