Fatal autopilot crash, NHTSA investigating...

[Don85D]

In my opinion the root cause of this collision is the failure to look far ahead on the part of all parties involved including AP. The fundamental first rule of driving that I have taught my children and now my grandchildren is that you need to look far far ahead. If you look far ahead your instinctive steering skills will take you where your eyes are looking, you will have time to pick an escape route in a crisis and most of all you will observer human behaviour which is adding risk. My risk assessment is continuous when I am driving. If I see a car waiting to make a move I know that the driver is calculating my approach speed and he may not get it right. If a car ahead is acting aggressively by tailgating or switching lanes to get a few car lengths ahead of the traffic I know that this behaviour is adding risk. If I see a car waiting to make a left turn in front of me I need to see their eyes to be sure that they see me.

When we first tested AP on a divided highway with tight curves it was obvious that the car was not looking far enough ahead for smooth steering inputs. To me AP was like slot cars from the 60s. Slot car pins in a track could keep your toy car in a lane but when speeds got high all bets were off.

Until cars can observe human behaviour at a distance to assess risk, drivers need to play their role and play it perfectly. Gadgets are distracting drivers and some are addictive reducing the urgency to continuously be looking far far ahead.

My thoughts for your consideration.

 

[SR22pilot]

If you add a radar at eye level the "is it a road sign?" ambiguity is eliminated.

Not necessarily. Remember you also have to tune it to pick up objects closer to the road. The most general case is looking for an object at standard bumper height. That is the most useful case to cover and the one you have to cover first.

 

[SR22pilot]

In my opinion the root cause of this collision is the failure to look far ahead on the part of all parties involved including AP. The fundamental first rule of driving that I have taught my children and now my grandchildren is that you need to look far far ahead. If you look far ahead your instinctive steering skills will take you where your eyes are looking, you will have time to pick an escape route in a crisis and most of all you will observer human behaviour which is adding risk. My risk assessment is continuous when I am driving. If I see a car waiting to make a move I know that the driver is calculating my approach speed and he may not get it right. If a car ahead is acting aggressively by tailgating or switching lanes to get a few car lengths ahead of the traffic I know that this behaviour is adding risk. If I see a car waiting to make a left turn in front of me I need to see their eyes to be sure that they see me.

When we first tested AP on a divided highway with tight curves it was obvious that the car was not looking far enough ahead for smooth steering inputs. To me AP was like slot cars from the 60s. Slot car pins in a track could keep your toy car in a lane but when speeds got high all bets were off.

Until cars can observe human behaviour at a distance to assess risk, drivers need to play their role and play it perfectly. Gadgets are distracting drivers and some are addictive reducing the urgency to continuously be looking far far ahead.

My thoughts for your consideration.

The radar looks ahead about 500'. Distance is a tradeoff with false positives. LIDAR and phased array radar can bypass this issue but those solutions have been expensive.

 

[kirkbauer]

Cruise Control Could Be Hazardous To Your Health

 

[ecarfan]

Well, obviously. I pointed this out yesterday in the unending NHTSA investigation thread. Quote from the article:

"Drivers relying on cruise control could be at a higher risk of accidents related to drowsiness or distraction, according to a recent study out of France that raises a caution flag about the auto industry’s rush to ease the burden of operating a car.

Most new cars and trucks now come with cruise control, which allows a car to automatically maintain a constant speed. Many newer vehicles have advanced cruise control systems that adjust the speed to keep a safe distance from the car ahead in traffic.

However, a recent study by the University of Strasbourg in eastern France has found that drivers’ vigilance – and their ability to respond to emergencies – is impaired when they are inactive for long periods and rely on the cruise control.

Based on a series of trials on simulators with a group of 90 drivers of varying ages, the researchers found that episodes of drowsiness increased by 25% when participants were using a cruise control, compared to test groups that didn’t use them."

 

[ZTrekus]

After an inquiry, certain recommendations should be made to ensure that similar incidents can never occur again. I can think of a few for starters:

1. Tractor trailers should not be painted white but rather a colour that is quite different and stands out from the environment;

2. Tractor trailers should have something hanging from their sides so as to either engage the sensors on cars or to prevent cars from wedging beneath them without the car crushing and activating the airbag;

3. Tractor trailers should be fitted with radar beacons which all cruise control raiders can detect;

 

[22522]

Not necessarily. Remember you also have to tune it to pick up objects closer to the road. The most general case is looking for an object at standard bumper height. That is the most useful case to cover and the one you have to cover first.

I should be more specific. This is an additional radar looking for reflections coming from objects at the same altitude/height. So there is no geometry to confuse the system.

Here is the logic.
The velocity vector of the vehicle lines up with the radar return from the head height radar. Stop the vehicle.

This is he same logic that puts the primary radar, used in traffic jams, at bumper height.

So this is a secondary radar to protect the windshield, and occupants who look through the windshield.

 

[22522]

After an inquiry, certain recommendations should be made to ensure that similar incidents can never occur again. I can think of a few for starters:

1. Tractor trailers should not be painted white but rather a colour that is quite different and stands out from the environment;

2. Tractor trailers should have something hanging from their sides so as to either engage the sensors on cars or to prevent cars from wedging beneath them without the car crushing and activating the airbag;

3. Tractor trailers should be fitted with radar beacons which all cruise control raiders can detect;

Please look. There is no balance in your suggestions.

 

[kirkbauer]

Yup, obvious, but the mainstream media seems to forget.

 

[nctalkinghead]

Just a curious question: If you are behind the wheel of an automobile and paying any attention at all, when was the last time you failed to see a 53' foot trailer - regardless of its color - and the tractor pulling it in front of you? Some of these suggestions are really way over the top!

RIP Josh

 

[Gynob001]

1. Tractor trailers should be fitted with transponders that identify their positions to others.
2. Tracter trailers should be the same size as passenger cars-at least the height.
3. They all should have glowing neon paint to be seen in the dark.
4. Tractor trailers should have car spotters/cow catchers in them.
5. Tractor trailers should be capable of lifting themselves up when a car tries to go under them.
6. They should run in their own track.
7. When a camera tries to detect them, they should "smile" and be visible.
8. They should take the old name-The Train or the locomotive.
9. Whe crossing a busy intersection, they should automatically blow horn/whistle loud enough to be heard above the noise of a video/audio player.

This is not intended to be a serious posting.

 

[David29]

And, as I think I said in the thread about the (now infamous) accident, what was at play and potentially insufficient in that accident was really the cruise control (TACC) functionality, not the steering control. Many brands of cars have similar functionality, so calling it the first accident with "autopilot" is most likely mis-characterizing it.

 

[Seesaw]

I can imagine Brown not hitting the brake because he believed the truck paused while blocking only the passing lane and Brown believed he was going to pass in front of it but the trucker "gave it the gas" (his words) and accelerated causing Brown to hit it before braking.

This reminds me of something that happened to me. I was driving along a rural road at about 100kmh. I approached a railway crossing (in rural Australia many crossing have no lights or crossing arms) and saw a freight train parked 60 meters away perpendicular to the road with a nearby rail maintenance 4wd ute. I thought "Ok its broken down and stopped and far enough away" that I am fine to whizz across the crossing.

Little did I know it wasnt actually stopped at all and was crawling at a very slow rate - probably due to engine problems. From my angle perpindicular to it and the high speed I was doing, coupled with how big the train is and how slow it was moving - I simply loss complete visual perspective of how much it was moving and that it was going to intersect me at the crossing. When I realised it was far too late to brake and so I gunned the accelerator but I missed it by mere meters (but felt like feet).

 

[FloridaP90D]

A 2012 youtube showing an Audi A6 did not brake for tested object of carton boxes wrapped with aluminum foil.

The tester had an overexpectation that the Adaptive Cruise Control should brake to a stop.

That system was imperfect for this test of crash avoidance but it has been on the market before classic non-autopilot Model S were produced.

I suspect other current systems are imperfect too.

They just call them "Assist" like Volvo Pilot Assist, Mercedes Driver Assistance Package... instead of "beta."

It's just another way saying that you are still responsible because the system just helps you.

The road to pass a DMV test is incremental.

Notice we are very near the bottom and no where near SAE level 5!

SAE automated vehicle classifications:

• Level 0: Automated system has no vehicle control, but may issue warnings.
• Level 1: Driver must be ready to take control at anytime. Automated system may include features such as Adaptive Cruise Control (ACC), Parking Assistance with automated steering, and Lane Keeping Assistance (LKA) Type II in any combination.
• Level 2: The driver is obliged to detect objects and events and respond if the automated system fails to respond properly. The automated system executes accelerating, braking, and steering. The automated system can deactivate immediately upon takeover by the driver.
• Level 3: Within known, limited environments (such as freeways), the driver can safely turn their attention away from driving tasks.
• Level 4: The automated system can control the vehicle in all but a few environments such as severe weather. The driver must enable the automated system only when it is safe to do so. When enabled, driver attention is not required.
• Level 5: Other than setting the destination and starting the system, no human intervention is required. The automatic system can drive to any location where it is legal to drive.

These classifications are helpful. I have autopilot on my vehicle, but do not use it, nor do I have experience with it beyond using it only twice.

Im wondering (if there is a better place for this pls advise) about a scenario here in FL: we have signs that tell bicycle riders to use the entire lane. Many do and travel side by side in the lane.

Would autopilot detect and warn if it sees bike riders? Would it turn off autopilot?

One other point: I recall CEO talking about travelling from one city to another without touching the wheel. This gave me and my wife the impression the caf drove itself. In light of recent news, I feel like this functionality is much more like cruise control (which i also do not use)

Thx

P90D in Miami

 

[DFibRL8R]

After an inquiry, certain recommendations should be made to ensure that similar incidents can never occur again. I can think of a few for starters:

1. Tractor trailers should not be painted white but rather a colour that is quite different and stands out from the environment;

2. Tractor trailers should have something hanging from their sides so as to either engage the sensors on cars or to prevent cars from wedging beneath them without the car crushing and activating the airbag;

3. Tractor trailers should be fitted with radar beacons which all cruise control raiders can detect;

I do find the issues identified by Auto Pilot interesting in terms of improving roadways (and yes even other vehicles) to make these systems work better. If more semi autonomous cars exist that rely on properly marked lanes to function correctly, shouldn't municipalities focus more on maintaining these markings? What about construction zones? Should these be logged into a mapping system that disables AP functionality based on GPS location until the road is returned to the proper standard? And why not require modest safety modifications to an 18 wheeler to improve viability to decrease the chance of a fatal collision. Even marketing safety equipment to cyclists that makes them more "visible" to a car with AP would be nice. Not arguing against driver attentiveness and improving hardware/software to decrease errors but at some point there are other solutions that can improve the safety in these situations.

 

[McRat]

Not paying attention and inadequate training are the 2 biggest killers in cars.

It does not matter what model car/truck/bike/plane/boat it is, or what it is equipped with, or is not equipped with.

There are very, very few 'accidents'. Lots of crashes though.

 

[Ludus]

The radar looks ahead about 500'. Distance is a tradeoff with false positives. LIDAR and phased array radar can bypass this issue but those solutions have been expensive.

It would be surprising if the radar data set from an overhead sign or bridge wasn't sufficiently distinct from the set for a trailer across the road that deep learning algorithms can't be trained to distinguish them with no change in sensor hardware. It seems likely the issue just never came up before. It would be necessary to train the system to distinguish signs and bridges from normal impediments because they'd obviously be putting up flags all the time. Further refining that to discriminate unusual cases like trailers that have enough clearance that the radar can see under them ought to be possible too.

 

[McRat]

Clarification on the 'accident' concept:

If you drive without the proper skills, the logical outcome is a collision. The illogical outcome would be safe operation in traffic.

So it would be an 'accident' if you DIDN'T crash if you don't pay attention or lack skills. Normally cause and effect says a crash is the correct outcome.

 

[Chopr147]

True , making the drivers job easier can result in lowering driver alertness. Which is why lawyers spends hours upon hours coming up with new ways to say "not our fault"
Side note: Talking to a friend yesterday about Tesla. His opinion was Tesla should not be allowed to even have the AP. I disagree with him but some people think we all need to be protected from ourselves and technology

 

[Norfolk Enchants]

I think the problem with that is the low clearance when going over large humps.

I love the idea though, and the first thing that comes to mind is the side fairings that many trucks use now to increase their aerodynamics. Maybe they should be a requirement. They would help improve fuel efficiency and protect other drivers with advanced braking systems. Since the fairings are much closer to the ground, the braking systems should detect them

Yes, I'm sure that low structure in-between the wheels could scrape or inhibit the truck in an extreme ramp break-over situation, but I'm not sure how many semis would ever get into a situation like that. Also, If this were an issue. I think that this would also be the case with the aero side fairings that you mention. Furthermore, the few US semi-trailers that you do see with under-ride rails and every EU semi, seem to manage fine traversing obstacles with low side protection rails. Unfortunately, I suspect it just comes down to cost for the trailer manufacturer

Great ides to mandate the aero fairings though!, would save fuel, emissions, and if they made them structural, lives in incidents such as this. How hard can it be for NHTSA to get this done!