Hate to say it but guys got a point

[Steve Ekstrand]

I keep looking for a car with gratuitous intended acceleration.

[Will Kalman]

Hilarious video that demonstrates how a properly functioning Prius works.

But what happens when the Prius malfunctions?
That shifter knob? Software controlled.
That power switch? Software controlled.
The brakes (via ABS unit)? Software controlled.

Now let's see how funny it is when a software glitch slams the throttle open, disables the brakes via the ABS unit, and ignores the shifter and power buttons.

You get to see the mighty full force of the Prius giving everything it's got, as it slowly comes to a stop from its dragging rear tires.
What's the E stand for in E-brake again?

My point was that this video (and all the other ones like it) only demonstrates a properly functioning car, not a malfunctioning one where the service brakes could conceivably be disabled.

E-brake effectiveness depends on both on the power of the engine and the selected gear. Most FWD cars will drag their rear wheels all day long at full-throttle in first gear and probably bog down in 3rd or higher. I think that would be a very interesting test to see.

That fact that these complaints are only happening in Toyotas somewhat rules out driver-error. I really do think it's an edge-case software bug.

[Steve Ekstrand]

The scammers are only targeting Toyota since a handful of real cases hit the media.

[John Stimson]

A software bug that sets full throttle, and sets ABS to ice mode, and then enters a condition that is unresponsive to additional inputs (such as the shift lever or the power button) seems pretty far fetched. That's three failures all at once. If that's really happening, you should be hearing a lot more reports of each failure type happening individually.

In response to the person who commented on Wozniak's cruise control discovery -- sure, it is a familiar experience to have your computer queue up button presses and store them until they have all been satisfied, but that does not make it a good user interface. Cruise control is controlling an analog system, and it should respond in real time by accelerating as long the "accelerate" button is pressed and not accelerating when the button is not pressed. The label on the button is "accelerate", not "increase speed by 1MPH". The brake pedal cancels cruise control anyway, so by itself, even a badly designed cruise control couldn't be the cause of this media frenzy.

[Kristoffer Gjevre]

...

In response to the person who commented on Wozniak's cruise control discovery -- sure, it is a familiar experience to have your computer queue up button presses and store them until they have all been satisfied, but that does not make it a good user interface. Cruise control is controlling an analog system, and it should respond in real time by accelerating as long the "accelerate" button is pressed and not accelerating when the button is not pressed. The label on the button is "accelerate", not "increase speed by 1MPH". The brake pedal cancels cruise control anyway, so by itself, even a badly designed cruise control couldn't be the cause of this media frenzy.

That is an interesting comment, because I almost exclusively tap the buttons to increase or decrease curise control speeds when I need to adjust it, and it has worked in all cars I have driven that has had cruise control, and BTW, the buttons in my car are marked - and + (not "accelerate") and the manual states:

Increasing Speed While Using Cruise Control
...
To increase your speed in very small amounts,
move the switch briefly to resume/accelerate.
Each time you do this, your vehicle will go
about 1 mph (1.6 km/h) faster.

Reducing Speed While Using Cruise Control
...
To slow down in very small amounts, briefly
press the set button. Each time you do this,
you will go about 1 mph (1.6 km/h) slower.

And I love the fact that I can tap the buttons 5 or 10 times times when I enter a different speed limit zone to adjust my speed accordingly.

[Kristoffer Gjevre]

http://jalopnik.com/5492199/exclusive-e ... -a-scammer

http://news.yahoo.com/s/ap/us_runaway_prius" onclick="window.open(this.href);return false;

Toyota dismisses account of runaway Prius
...
In Sikes' case, Toyota said it found he rapidly pressed the gas and brakes back and forth 250 times, the maximum amount of data that the car's self-diagnostic system can collect.
...
Toyota spokesman John Hanson said the event data recorder — a car's version of the "black box" inspected after plane crashes — would be of no use to investigators because it only stores information when the airbags are deployed. The box only stores four to six seconds of information before the airbags go off, he said.

[Kristoffer Gjevre]

Another unintended acceleration case in Norway where supposedly the accelerator pedal got stuck (floor mats are being blamed):
http://www.vg.no/bil-og-motor/artikkel.php?artid=594219" onclick="window.open(this.href);return false;

The driver is in the hospital, but the son said that his father tried to brake, press the stop button, and to put it in neutral with no success...

Ifølge sønnen Mergim Hoti prøvde faren å stanse den ville ferden med å bremse, trykke inn stoppknappen og sette giret i fri, uten å lykkes.

In the same story a person that rented a Prius (and drove to my hometown) said the accelerator pedal got stuck 3 times and he kicked it each time to get it to work again:

En av disse er Sune Adolfsson, som forteller at gassen i en Toyota Prius leiebil hengte seg opp tre ganger på en tur fra Oslo til Fagernes tidligere denne måneden.

Gassen hang helt til jeg sparket gjentatte ganger i pedalen. Opphengingen varte i flere sekunder hver gang, forteller Adolfsson til VG Nett.

BTW, a few years ago there were some crazy stuff happening to a few, C6 (or was it C5?) Corvettes with bad steering censor connectors that caused the car to lock up one of the front wheel brakes, one was even at an autox... GM put out a service bulletin on it...
Drive by wire and electronic nannies are great...

[Will Kalman]

A software bug that sets full throttle, and sets ABS to ice mode, and then enters a condition that is unresponsive to additional inputs (such as the shift lever or the power button) seems pretty far fetched. That's three failures all at once. If that's really happening, you should be hearing a lot more reports of each failure type happening individually.

No, that's one failure. Of the ECU. I don't see how you can call three I/O registers in a single controller separate systems. And when the system enters an undefined state, you certainly can get undefined outputs.

Would you bet that when your computer crashes that there were no spurious memory or electronic logic flips? Not one funky byte went down the USB bus? You've never had a peripheral go unresponsive or have your printer blurt out reams of unintelligible code because of an I/O error or driver freak-out? Not one internal logic line flipped to an unpredictable state? Every CPU made comes with a sheet of errata that basically states "we thought we programmed it to X, but it does Y". Who says there isn't an extremely rare edge case in the software (or even in the hardware/microcode) that sends max values (i.e. "all ones" - 0xFF) to the throttle plate servo controller and the ABS system just before it crashes and becomes unresponsive to further input?

Aside from all that conjecture, my whole point is that it's useless to illustrate the behavior of a malfunctioning Prius by using a functional one. It's like calling tech support because your CD tray won't open and they say "look, you idiot, you just press the button on the front and the tray pops out, mine works just fine all day - click - bzzzz, click - bzzzz, click - bzzzz. Learn how to use a computer you moron!"

[Marshall Grice]

No, that's one failure. Of the ECU.

you're assuming all the i/o's are controlled by one processor. generally the throttle is not ecu controlled (stand alone circuit even though it technically is inside the same housing as the ECU), nor is the abs(separate processor). Yes they both take inputs from other processors but they by themselves aren't tied to the processor functions of the main ecu and thus aren't (or shouldn't be) linked to ecu lockups. Plus if your ecu locks up the engine should quit immediately...among other things.

[Will Kalman]

True, there are more parts than that. But a sub processor can send out an errant command to another and then get into a state where it thinks things are running smoothly. Just as you can have your computer merrily playing an MP3 in the background (so you know the CPU is working) but the UI is hung. It is possible for some functions to end up in an undefined state while others continue working. It's very likely that there is a sub processor running the engine just fine but the "main" one that coordinates them is hung or sending out incorrect instructions (i.e. the last thing the main ECU said was "full throttle and we're on ice"). Even a singular CPU can continue running the engine based on interrupts while the main routine has hung. Heck, there may be a processor dedicated just to handling user input (shifter, gas pedal position, power button status, etc) that has an issue and sends erroneous commands to the ECU.

Ask anyone who's written a muti-tasking or multi-threaded application how hard debugging can be when things operate independently.

[John Stimson]

You think that there's no watchdog process that senses when a critical control process (like one that has the final say over both the throttle position and the ABS mode) has stopped or entered a non-operational loop?

[Will Kalman]

You think that there's no watchdog process that senses when a critical control process (like one that has the final say over both the throttle position and the ABS mode) has stopped or entered a non-operational loop?

We're picking apart the conjecture on the cause (and none of us has the real info on the Toyota control system). I'm only talking about the logical fallacy of the video.

[Theo O.]

...Ask anyone who's written a muti-tasking or multi-threaded application how hard debugging can be when things operate independently.

I do and yes, debugging can be tricky. However, muti-threading makes detection (of a problem...not necessarily the specifics) even easier. Each thread can report it's status, if I get a funky response or no response from a thread then I know something is up. In which case the last thing I would do is send the throttle into WOT and disbale the shift mechanism.

Looks like the San Diego event is a fraud anyways... http://online.wsj.com/article/SB2000142 ... 42938.html

I TOTALLY agree that it's not convincing to use the result of a functioning car to dismiss issues with a malfunctioning car. But given same hardware/software and no failure detected in any hardware/software, how do you determine that the car is even malfunctioning...i don't know anymore

[Kristoffer Gjevre]

I TOTALLY agree that it's not convincing to use the result of a functioning car to dismiss issues with a malfunctioning car.

Or, like a computer that has been rebooted, everything works fine after the issue happened and the car was turned off...
I know our company spends a lot of time trying to reproduce bugs customers report and sometimes can not and have to work remotely on the customers machines.
The only trace/proof of what happened is in whatever logs was produced (if you are lucky) and possibly data altered.

[Will Kalman]

I do and yes, debugging can be tricky. However, muti-threading makes detection (of a problem...not necessarily the specifics) even easier. Each thread can report it's status, if I get a funky response or no response from a thread then I know something is up. In which case the last thing I would do is send the throttle into WOT and disbale the shift mechanism.

You're right, I'm not talking about just problem detection, I'm talking about determining cause (and thus a fix). Race-conditions are the classic nightmare of multi-threaded (NOT so much multi-process) programs. And I'm not saying the code as written would send WOT, I'm saying an undefined condition might. Take buffer overflows - it's the data fed to the routine that creates NEW code to be executed. Maybe there's an overflow somewhere or a shared memory space with stale data in it that contains just the right info to send WOT before encountering the next piece of data that's crash-wothy. Maybe, just maybe, there's a flaw in the Toyota software that only happens once in a million or billion times contains just the right conditions for just the right code to be in the wrong place at the wrong time.

[Steve Ekstrand]

I heard that if you start the Prius then press the horn three times, depress the brake pedal once, set the radio to KUSC, and then turn on the a/c, all with the right blinker flashing, the car absolutely won't accelerate as intended.

[Theo O.]

I heard that if you start the Prius then press the horn three times, depress the brake pedal once, set the radio to KUSC, and then turn on the a/c, all with the right blinker flashing, the car absolutely won't accelerate as intended.

Correct. You would have drained the battery by them

[George Schilling]

I heard that if you start the Prius then press the horn three times, depress the brake pedal once, set the radio to KUSC, and then turn on the a/c, all with the right blinker flashing, the car absolutely won't accelerate as intended.

You forgot the part where you put your thumb in your mouth and blow real hard.

[Dan Shaw]

Dood, it's all about cheat codes...plug in a USB controller to the MP3 adapter port and hack away.