Slashdot Mirror
Tesla: Model X Accident Caused By Driver Error, Not Autopilot (computerworld.com)
An anonymous reader writes: Tesla has responded to a recent report from a Model X owner claiming their vehicle suddenly accelerated at "maximum speed" by itself, jumped a curb and slammed into the side of a building while his wife was sitting behind the wheel. They said it analyzed vehicle logs, "which confirm that this Model X was operating correctly under manual control and was never in Autopilot or cruise control at the time of the incident or in the minutes before. Data shows that the vehicle was traveling at 6 mph when the accelerator pedal was abruptly increased to 100%. Consistent with the driver's action, the vehicle applied torque and accelerated as instructed. Safety is the top priority at Tesla and we engineer and build our cars with this foremost in mind. We are pleased that the driver is ok and ask our customers to exercise safe behavior when using our vehicles." When will people stop lying about Tesla's Autopilot mode crashing their cars? One Tesla owner recently filed a Lemon Law claim against the company over a high number of quality control issues.
Majority voting with three systems has turned out to be dangerous on quite a few occasions. At least with two systems disagreeing, you can decide "I don't know which one is right, so I'll do the safe thing". With two out of three giving an incorrect value, it leads to misplaced confidence in this wrong value.
Examples:
- (human): Primary altimeters different between captain and F/O, standby altimeter agrees with captain, but F/O altimeter was correct. Don't remember the flight number or even the company, but they crashed into a mountain. If they hadn't had the standby altimeter, they would have gone with the lowest of the two indications until they could intercept a glide slope somewhere.
- (automatic): The A320 has independent Angle Of Attack probes, but if you climb through heavy icing, it turns out two can freeze up at the same time. The system sees a confirmed stall (even though it should be impossible given the current speed, attitude and inertial measurements) because two systems agreeing can't ever be wrong, and it pushes the nose down. Even if both pilots pull back on the stick, the nose continues to drop. We now have an official procedure to shut off two of the Air Data Reference units, leaving only one remaining. That's the only way of shaking the absolute confidence of the system in two identical but false signals.
Tesla lost because the judge determined the show was entertainment and not a documentary or news program and had no requirement to be factual.
They lied and faked almost the entire thing.
I did have an issue with an Audi several years ago. My cruise control would periodically turn itself on. They checked the computer, and it was reading occasional faults in the cruise control switch, so they replaced it. The problem continued. Sometimes over bumps, but also sometimes when I would turn on my turn signal. The turn signal and cruise control switches were both stalks on the left side of the column. The dealer read the codes again, and again said the cruise control switch was reporting faults and wanted to change it again. When I protested, and pointed out that it would turn on when I turned on my turn signal, they assumed that I was mistaking the two stalks. (to which point I asked them to investigate why my turn signal was turning on when I activated my cruise control...) It was only after the service manager finally driving it around and being able to duplicate the issue that they believed me. They ultimately replaced the control module in the steering column, which contained both switches, which solved the problem. Not saying that the brakes and gas pedal go into the same control module, but I can certainly see a case where the control module has a fault which is reading the wrong input.