YIKES: Experienced unexplained acceleration, already had brakes reprogrammed

Very cool data, and thanks for the effort in acquiring and posting this.

 

Minor quibble, Bob. The radial acceleration is a = w**2*r, and w = v/r, so a = v*v/r, where r is the radius of the curve and v is the velocity.

So, I looked the location up in Google. My "paper on the screen" gives a radius of the turn of about 42.5 m; 25 mph is around 11.2 m/s. So, when you're in the turn the acceleration works out to be around 2.9m/s**2, which is about 1/3 G.

Of course, unless you're going round and round in circles, nobody jerks a car so the centripetal acceleration goes from 0 to 1/3G instantly, so one would expect the G forces to rise until one is truly in the curve, then decrease as one comes out. Not that I'm doubting your results (far from it!), but that throws some junk into the analysis.

Another interesting point, now. 0-60 acceleration time on a 2010 Prius is listed as 9.8 s. Well, a = (v-vo)/(delta t), which, since 60 mph is 26.8 m/s, gives an acceleration of 2.73 m/s**2.

Divide that by 9.8 m/s**2 (acceleration of gravity) and we've got about 0.28 G. Now, I know that the acceleration peters out as one goes faster, but I would've guessed that at 25 mph one would still be on the steep portion of the acceleration curve. However, your plots are showing an x-axis acceleration, whether on the straightaway or in the curve, of little less than 0.1 G. So: When you kicked the thing into "resume", how did it feel to you? A neck-snapper, or something more sedate?

Finally: So, having read the O.P.'s comments, I sat in the car with my foot on the gas pedal and moved the wheel and my knee around. Yep, there's interference. However, in order to get that switched moved around (or depressed, since the button's on the end) I really had to stick it into my leg; it wasn't a feather brush feeling, more like being poked with a stick.

So, now I'm wondering, again, about software. The cruise control is in the ECU. So, as a WAG, there's a switch or case statement in the code. At around 25 mph there would seem to be some interesting bits of code:

1. As the speed is going from below 25 mph to above 25 mph, checking to see if the Cruise Control (CC) resume is on; if so, engaging CC.
2. As the speed is going from above 25 mph to below 25 mph, checking to see if the CC is engaged; if so, disabling same.
3. As the speed is going from above 25 mph to below, if the CC is "on", disabling the use of the Resume function.

It strikes me at that speed the ECU code would be, well, relatively busy with lots of "if, else-if" statements in the code. Further, just to throw junk on the fire, all measurements taken from the real world are going to be noisy. Analog or digital filtering helps.. But if somebody missed a state machine input or output, that's where I'd put the staring eyeballs.

Nasty thought: The OP said that he was taking that turn in the rain. Not fast, but suppose he went over a manhole cover and a wheel, well, skidded a bit? Just what does the Prius use for speed measurement? A couple of my old cars had a speedometer cable threaded through a front wheel. I imagine that the technology in the Prius is past that.. But if they're measuring speed but looking at magnets on an axle, that, too, might show a fast-moving blip in velocity at, maybe, just the wrong time in the code.

Great. We'd have to look at the code to see if there's a trap in there, and that's not going to happen. Oh, well.

KBeck

 

Ordinarily I use "F=ma" or "F/m = a" so either form of the equation works but I'm not parochial about the data:

DATA-004.CSV - 1.3 MB
DATA-005.CSV - 1.3 MB
DATA-006.CSV - 1.3 MB
DATA-007.CSV - 380 KB
​

Each file consist of 64,000 samples so to put the data into an excel plot format, I averaged 5 points for each plotted data value. Not shown are some maximum acceleration events at the end. If you are interested in copies of the data, send me a PM with an e-mail address and I'll gladly forward the original data.

In my data analysis, I used a simple offset to handle mounting errors (aka., not using a spirit level to ensure a flat mounting. A more sophisticated, vector based correction might shift the values enough to match your expected values. But I haven't tried to calculate the expected values based upon geometry and speeds.

Not shown, I also did a maximum acceleration test on the flat and in the curve. This was 'fun' but not relevant to the hypothesis of inadvertent cruise control activation.

Mostly I wanted to present a model, an approach that the OP might use next time they are in the area. If they can replicate what they remembered using any of the insights provided, good news. If not, that would be good to know too. Regardless, right hand turns can swing the stalk down and it may or may not be felt. However, it can activate a latent cruise control value. In a curve, this can rapidly increase the perception of a force or acceleration.

Bob Wilson

 

First I want to thank you for doing the analysis. This is how we collaborate in engineering and science:

This is an excellent approach so I substituted PowerPoint as my electronic paper. PowerPoint has the ability to:

• read out object dimensions - normally using inches in the USA
• resize objects maintaining aspect ratio - so I could resize a screen shot of the curve to fit in a PowerPoint chart
• resize graphics - so I could make a circle overlay and a scale line (see lower left)

My measurements show the circle radius to be ~69.65 m. When I do the math, the acceleration comes out at ~0.203 G.


So when I look at the first two laps through the curve:

I'm seeing closer to .166 G, close:

(.166 - .203) / .166 = ~22.3% error
​

Now there is a crown on the road the would tilt the car several degrees and there may be a shallow bank but I made no attempt to measure. Both the crown and any banking would reduce the measured lateral G and slightly increase the vertical G. I looked at the data again to see if I can derive any change in the medians:

	Column 1	Column 2	Column 3	Column 4	Column 5	Column 6
0	Ax	Ay	Az	segment
1	77	124	1015	curve median	1	542 samples
2	64	-60	1012	straight median	1	082 samples

1012 counts ~= 1G.
​

Based upon this data, the curve measured (124 - (-60)) / 1012 ~= 0.182 G. Then looking at the change in G, (1015-1012)/1012 ~= 0.002964. So the angular change seems pretty slight. However, this gives a more accurate reading of the lateral acceleration of the curve, 0.182 G (184/1012). But there is a bigger error from the speedometer.

We know the ZVW30 speed indicator at higher speeds shows 1 mph higher than the GPS measured and ground-based, clock measured values, again giving a slightly lower true speed and lower G. So if the true speed is 24 mph from an indicated 25 mph:

• 24 mph -> 10.73 m/sec
• (10.73 * 10.73) / 69.65 m -> 1.65 m/(s*s)
• 1.65 / 9.8 -> 0.168 G vs. 0.166 G
• (0.166 - 0.168) / 0.166 ~= 1.2% error from 1 mph error

So a speedometer error of just 1 mph would be enough to account for the calculated versus measured error. This also gives an idea of how the any acceleration in the curve changes the acceleration or perceived force.

In measuring, there will always be some degree of observer error. As an undergraduate, we learned about 'engineering error' which seemed to let us magically achieve a desired answer. But in the real world, we can get multiple answers using parallel approaches:

	Column 1	Column 2	Column 3
0	circular acceleration	method
1	0.203 G	calculated at 25 mph
2	0.182 G	first	measured G
3	0.166 G	eyeballed	G of curve from graph
4	0.165 G	calculated at 24 mph

from this posting

Still, these seem pretty close together and I would probably use, 0.18 +/- 0.017 G, as my best guess of the true value.

Thanks because this shows how process, the way we approach a science or engineering problem, is the key. We may initially get different numbers but as we refine our metrics, we come closer and closer to the true value.

Bob Wilson

 

Bob,

You're right about the collaboration! Well, if you can get to 1% error on something like this you've got my admiration! Wish I'd thought of your Powerpoint trick, that's better. I think that had I used the larger (correct) radius my original analysis would have been within 10% or so, and the error bars on the speedometer would have taken care of the rest.

I think you've proven one thing: Once the CC gets set to 65 mph or something, it stays there so long as the CC or car doesn't get turned off. Umm.. I think you said that. That's right, right?

I guess the question is, now, does the data you've collected and that we've analyzed support or not support any hypotheses? And, while mumbling about hypotheses, are there any other tests that might push one or the other hypotheses in one direction or another?

Hypothesis #1: Strange gas pedal issues. Doesn't seem likely, the OP didn't have his foot on the gas. I suppose the thing could have failed just sitting there... (thinking back to discussions about the electronics, that is.) I don't like the idea. Comments?

Hypothesis #2: The right turn that the OP took, with his foot off the gas and flat on the floor, was enough for the steering wheel to curve over and his leg and the CC to meet with each other.

Hypotheses #3: Some bizarre software bug in the ECU.

By now people have got me pegged as a doom-and-gloom type, but there's a couple of reasons why I find this one interesting:

1. Software reuse. Somehow, if I was writing software subroutines to control engine throttle on one car, I'd kind of expect that that software, once vetted, would see reuse. Hey, why write new, complex code, when, sitting right over there, is an already tested package that's Known Working? Not that there's been many, or any, other verified Prius runaways, but there sure have been a lot of other Toyota fly-by-wire cars that have had suspicious (not proven!) incidents that might be tracked back to a funny ECU. Having a throttle override (I hate calling it a brake override, it's not the brakes being over ridden!) keeps Priuses from smashing brick walls, so maybe people have seen this and not bothered to report.
2. There's a lot going on around 25 mph in the software. Further, the world is full of noisy sensors, and that includes the speedometer data. Good engineers would filter the living junk out of any raw data, but I wonder: Put enough lines of code in there and, the more code, the more likely something is Not Quite Right. Especially if the code is running through lots of case statements.
3. If something like this crops up, the software could do anything. Full throttle? You betcha. Somewhere in between? Makes sense, too.

Now I'm getting crazed.
Hypothesis #4: The CC was off, but when you turned into the curve the CC button got banged and (boy, this is nuts) somehow jammed the CC stalk so it was jammed straight up. Sticky CC stalk, anyone?

Questions:
1. Mr. Prius in PA: Have you traveled back to the site? Does going around the curve line the CC up with or near your leg? (That is: How far was the steering wheel canted over?)
2. Presumably you had your foot on the gas, then moved it to the floor. Is it possible that, during that motion, you hit the CC knob?
(Mind you, I tried doing that. My feeling is that I would have had to be really distracted in order not to notice the hit on the knee.) Were you in a deep conversation with your significant other when it happened?
3. Bob, Mr. Prius in PA: Bob's results show that, when he hit the resume button, the car went to roughly 1/2 of max acceleration. That's not full throttle. Comments, Bob? That, by the by, also seems to be what I see when, after going through a toll booth at 25 mph or so, what I get when I kick in resume back to 65. It's doesn't try to screech the wheels, not that one can do that with a Prius, but it does pick up speed. That also makes sense from Toyota's point of view: full throttle anything when "resuming" doesn't sound quite right. Mr. Prius in PA: Did it seem like full throttle to you? Or did it seem something less?
4. OK, this is crazed. Every drop a cup of coffee on the steering wheel? Did it get cleaned up, but the CC stalk hole has some sticky stuff in there? (yeah, yeah, I know. But I gotta ask.)

OK. That's four hypotheses. Anybody got any other ideas? How about any more tests? Or any better questions?


Back to schematic capture.

KBeck.

 

Patience is one of the hardest things about intermittent problems. Worse, they often come with at least one other, independent cause.

We need to wait for the next incident and again, treat each posting with respectful, quisitive, curiosity. Someone who posts here has something we all lack, direct experience. So let's try and understand every symptom, as much detail as possible. Then try the hard stuff, to replicate their experience and understand what is going on.

Thanks,
Bob Wilson

 

I really love your shark analogy..Its spot on...But one problem, shouldn't the swimmers be informed at the time and place where the shark attack took place ? So they can make their own decision whether they want to enter the water or not or maybe the authorities should tell them 4 months later of the matter ?

Reminds me of a movie...Where a girl gets killed by a shark attack , they analyze that's it a probable shark attack,they sweep the matter under the carpet and keep the beach open for fear of losing allot of business during their best holiday stretch, till someone else gets killed and the Police Chief gets slapped...Sound familiar?

 

We absolutely should be informed that sharks bite and what part of the country they are likely to bite or not likely to bite, and some will feel they need equally detailed information on how to use CC.

 

You're putting yourself through all this trouble when the culprit is most likely accidental activation of the CC stalk (since you admitted having the cruise control activated at the time) and/or the myriad of possibilities due to loss of traction (since it was raining at the time)?

 

Prius in PA, you are the only one who can really determine if it was accidental acceleration or sudden unintended acceleration. I for one believe that if there was a glitch in the mechanics or software, others would have experienced this problem already.

 

This is weird and unnecessary. Talk about poor customer dealing! You have my sympathies.

Was this just in response to your earlier report?

Did your NHTSA report include the VIN number? (The only way I suspect they may have flagged the car.)

Bob Wilson

 

I just find it funny that two engineers can't possibly understand that the biggest potential culprits are the accidental activation of the cruise control and loss of traction secondary to the rain, and are instead going to the extreme length to fault-isolate other less likely culprits.

 

They will Fault-Isolate or else the lawsuits keep popping up. Owners will not be happy with a simple explanation like floor mats or accidental CC activation.

 

Occam's razor:

• File an NHTSA report with VIN - the complainer's next visit to a Toyota service center results in the vehicle being held and a substitute offered. Hope you are ready for the surprise.
• NHTSA sees Toyota addressing their VIN reports - NHTSA is happy as Toyota is giving their reports merit.
• Consumer Reports and other 'safety advocates' - are thrilled, initially, as the action they so dearly want is being followed.

So what happens next?

1. Word of mouth - file an NHTSA report with VIN and your vehicle is 'taken' for diagnosis.
2. NHTSA slowly realizes they don't have the budget and are reduced to passive watcher, the users are making the reports and NHTSA have no 'value added engineering.'
3. Consumer Report and other 'safety advocates' notice (if they are clever enough) that the numbers of reports are falling off . . . the new program 'kills their golden goose.'

Bob Wilson
(the paranoid self)

 

I am curious, how can this happen unless the CC is turned on manually when the car is started or soon afterwards?

 

The OP admitted to having the cruise control on (see post #43) and it happened on a curve (see post #52).

 

Thanks, I missed those
I will have to see if my knee can do the same. But, in the meantime, not knowing her habits how far the foot is pulled off the gas pedal, this may not be the cause but the multiple wheel turns to the right is the direction to engage accidentally. this is then a design flaw.
ps.
Just sat in car. I need to pull my legs uncomfortably up under the steering wheel to make contact while turning the wheel. Perhaps that is what the OP's habit is going around a corner?