Tesla Driver Watching Movie While Using Autopilot Crashes Into Cop Car Because Autopilot Doesn't Really Drive, Dummy
Somehow Autopilot not only steered the car off the road, but directly into a parked police car with lights on. This is not the sort of mistake most human drivers would make. If the driver was using Autopilot properly—as in paying attention and not watching a ####ing movie while he’s supposed to be driving—this would not have happened.
Ok, so obviously the guy shouldn't have been watching a movie. But the bigger problem is that the car drove straight into a parked police cruiser with flashing lights. What kind of programming is going on there that makes the vehicle do that? I can't even get my head around it. Did it think the lights were a flashing portal the enter? Why would it aim for an object? If it didn't see it at all, then why would it leave the lane it was traveling in?
Ok, so obviously the guy shouldn't have been watching a movie. But the bigger problem is that the car drove straight into a parked police cruiser with flashing lights. What kind of programming is going on there that makes the vehicle do that? I can't even get my head around it. Did it think the lights were a flashing portal the enter? Why would it aim for an object? If it didn't see it at all, then why would it leave the lane it was traveling in?
The adaptive cruise in my E550 would do the same thing. The problem is these systems are able to speed match or adapt only to moving vehicles. The way it identifies them as vehicles is down to size and the fact that they are moving in the first place. When a vehicle is stopped, the car sees it as a stationary object that you are about to drive into and instead will prepare to emergency brake if you don’t start braking yourself.
On its face, it’s kind of silly and results in the following situations:
1. You are driving down a highway, there is a vehicle in front of you. You round a bend, and approach a lighted intersection that is red in your direction. The vehicle in front of you slows down and comes to a complete stop. The system detects the vehicle in front doing this, and also slows down and comes to a complete stop.
2. You are driving down a highway, there is no one in range of the system in front of you. You round a bend, and approach a lighted intersection that is red in your direction with vehicles already stopped at it. The car senses this as driving towards a stationary object, and instead of slowing you down as well, it continues on at your regular speed until it senses you are not intervening. The car’s pre-safe brake system will engage and emergency brake before impact. This generally doesn’t prevent a crash but rather aims to reduce the impact severity. You don’t take action and you rear-end the stopped car at a speed less severe than without the system, but a collision still occurs.
There are probably elements of #2 in the Tesla story. And while #2 seems like a horrendous oversight, I have found the car starts warning of (not braking though) impending crashes when I aggressively take a tight curve where there is a safety barrier immediately beside the lane. And really, these systems aren’t meant to be a complete replacement of driver attention and intervention.
I don't know how the Tesla's systems are trained to interpret flashing police lights but it might have been unable to process where it was because of them and strayed out of the proper lane.
I wish they didn't call it Autopilot. Even if it's 99.9% safe, there are still obviously scenarios it can't handle. The driver needs to pay attention and be ready to act when they arise.
I think it's just an example of how far off these systems are from being 100% autonomous, despite what Elon says. No safety system should steer into a stationary object. This Tesla would have had to depart it's lane of travel. If you were behind the wheel, paying attention and had autopilot on, and you noticed the car veering into a parked police cruiser with flashing lights, would that not give you pause about ever using this "safety" feature again?
Even autopilot on an aircraft won't land the damn plane for you. People are expecting too much, and perhaps some of that blame needs to fall on Tesla for touting the system in the way they do.
Even autopilot on an aircraft won't land the damn plane for you. People are expecting too much, and perhaps some of that blame needs to fall on Tesla for touting the system in the way they do.
It actually will.
Anyway, my point was it shouldn't actively steer you out of the lane into a hazard. That's a major flaw. That's beta test failure, not release to public.
Autopilot with takeoff and landing is probably a lot easier to implement than auto-driving. Far fewer variables.
A plane landing on autopilot in a snow storm is probably safer than a pilot doing it, given all the sensors and instruments, whereas a self driving car in a snowstorm is unlikely to stay out of a ditch for more than 15 seconds. Certainly a Tesla. Maybe with Lidar systems it becomes more possible. I think Volvo has done a lot of work with them, which makes sense given their headquarter location.
The Following User Says Thank You to Fuzz For This Useful Post:
What really need to happen is networking every vehicle together. If a car encounters a hazard it can tell the others around it and they can act accordingly. Imagine a child runs onto the road and the car can’t brake in time but it can swerve, but it needs the car beside it to maneuver to give it room. Coordinated vehicles will save lives, not to mention time and energy.
Even autopilot on an aircraft won't land the damn plane for you. People are expecting too much, and perhaps some of that blame needs to fall on Tesla for touting the system in the way they do.
Flew from Danriga to Belize City in the co-pilot seat. Pilot turned the plane onto the runway and never touched the controls again until turning the plane off the runway in Belize City. Only a 20 minute or so flight but that was quite nerve racking actually.
Flew from Danriga to Belize City in the co-pilot seat. Pilot turned the plane onto the runway and never touched the controls again until turning the plane off the runway in Belize City. Only a 20 minute or so flight but that was quite nerve racking actually.
What really need to happen is networking every vehicle together. If a car encounters a hazard it can tell the others around it and they can act accordingly. Imagine a child runs onto the road and the car can’t brake in time but it can swerve, but it needs the car beside it to maneuver to give it room. Coordinated vehicles will save lives, not to mention time and energy.
I think the new VW Golf has that.
Yep, from wiki:
Quote:
The Mk8 is the first Volkswagen vehicle to use Car2X, where information can be exchanged with other cars and the traffic infrastructure within a radius of up to 800 m (2,625 ft).
Quote:
Vehicle-to-everything (V2X) communication is the passing of information from a vehicle to any entity that may affect the vehicle, and vice versa. It is a vehicular communication system that incorporates other more specific types of communication as V2I (vehicle-to-infrastructure), V2N (vehicle-to-network), V2V (vehicle-to-vehicle), V2P (vehicle-to-pedestrian), V2D (vehicle-to-device) and V2G (vehicle-to-grid).
There is lots of talk that Tesla's autonomous system is their ace in the hole, that no other manufacturer can compete. It doesn't seem that is true at all. GM's supercruise takes the win.
Quote:
Even after two years, Cadillac’s Super Cruise remained our top-rated system because, when turned on, it uses direct driver monitoring to warn drivers that appear to have stopped paying attention to the road. General Motors told CR that Super Cruise will be on 22 GM vehicles by 2023. Of the other systems we tested, we saw minor improvements in lane keeping performance for the Tesla and Volvo. Systems that didn’t give clear warnings to the driver to pay attention, such as Volvo's, or that failed to keep the vehicle within its lane, even on fairly straight roads, such as systems from Buick, Mazda, and Land Rover, didn’t fare well in our overall scoring. When CR requested comment from these companies, several said their lane keeping assist systems were not designed to keep the vehicles centered. (See rankings below.)
There is lots of talk that Tesla's autonomous system is their ace int he hole, that no other manufacturer can compete. It doesn't seem that is true at all. GM's supercruise takes the win.
Tesla still ranks highest on capabilities and performance, but only by one point. It falls behind on monitoring the driver for awareness.
I'm probably in the minority of Tesla owners on this one but I wish they would just focus on making fun electric cars instead of pretending like the car drives itself... It doesn't, and it won't be good enough for years to come.
The Following User Says Thank You to Ahuch For This Useful Post:
There is lots of talk that Tesla's autonomous system is their ace in the hole, that no other manufacturer can compete. It doesn't seem that is true at all. GM's supercruise takes the win.
Tesla still ranks highest on capabilities and performance, but only by one point. It falls behind on monitoring the driver for awareness.
I guess it depends on what you're looking for. The ability of the car to autonomously pilot itself through dynamic traffic situations is what most people tend to be interested in because it involves so many variables. "Is the driver paying attention?" seems less impressive as a feature, although I don't think it needs to be stated how important it is to actually rolling out autonomous driving broadly.
__________________
-James
GO FLAMES GO.
The Following User Says Thank You to TorqueDog For This Useful Post:
I'm probably in the minority of Tesla owners on this one but I wish they would just focus on making fun electric cars instead of pretending like the car drives itself... It doesn't, and it won't be good enough for years to come.
Agreed. The focus should be on better battery life and making the cars more affordable to get more of them on the road. Autonomous driving is cool but not a necessity.
