Cruise Control feels like ICE

[S-2000 Roadster]

Rather than continue to dilute other threads, I wanted to discuss the quality and implementation of the Roadster cruise control.

Does anyone know whether Tesla Motors licensed or otherwise purchased some existing generic cruise control module (*), or if they implemented their own? The reason I ask is that cruise control in every ICE car I've ever tried has a really distinctive feel, and they all seem rather stupidly aggressive. My Roadster 2.5 feels just the same. Considering how smooth Tesla Motors managed to get all the other features of the Roadster to work (other than really slow, uphill reverse), I'm quite surprised that the cruise control does not feel a lot smoother.

Any comments?

I'm fairly new to the concept of automobile design, especially with regard to testing, standards, and any sort of mandatory approval bodies. I have this suspicion that federal laws mandate a great deal of minutia, and thus I imagine that licensing existing, approved technology might be far easier than jumping through all of the hoops and over the various hurdles. But I realize that I could be making this all up in my head. If anyone knows about these sorts of legal standards, I'd appreciate a reference.

[cinergi]

Fascinating ... I find the cruise control to be perfect. It's extremely precise, which is what I like. I don't feel anything odd while it's on -- I just go exactly the speed I set it at, no matter what.

[Doug_G]

I'm with cinergi. You set a speed and it stays at that speed. It smoothly increases and decreases power as needed, and you hardly notice it.

[hcsharp]

Agree with Cinergi and Doug_G. I don't think they used an existing module. I can't seem to get it to "coast" but maybe I haven't found that button. Or technique.

As a tournament waterskier I watched while several companies developed speed controls for tournament ski boats. They are much more accurate than automotive cruise controls. I can drive my boat through a slalom course (almost 1/4 mile) and the time will rarely vary by more than 3 thousands of a second regardless of how much the pull on the boat varies.

[kgb]

I agree with everyone else above. Since we have a cruise control thread, I've been meaning to ask. Anyone notice this:

Driving with cruise control on, but I desire to speed up a hair to get around someone. If I lightly touch the torque pedal, I seem to engage regen, because the car seems to slow. So I have to press past the "sweet spot" to get the car to speed up. This seems counter intuitive.

[S-2000 Roadster]

Maybe it's the difference between "driving" or "being driven," but I find my own acceleration techniques to be much smoother than the cruise control. Frankly, I think that I drive really fast and aggressive, but I've been complemented on how "smooth" I drive (that was in the CRX, though, long before this Roadster 2.5).

[NigelM]

I've had my Roadster 3 months now and only used the cruise control once, just to test it. Driving the Tesla is too much fun to want to use the cruise control and I find it much easier to maintain a constant speed in my Roadster than in any ICE I've driven. Although it's pretty flat around here - there's no such thing as Florida mountain rescue....

[vfx]

I just love how it LOCKS. I had not driven any other car that would stay at the same speed up and down very steep hills like the Roadster does. A nearby mountain pass is a favorite place for radar cops with several hiding places plucking cars off all day. For that stretch I now set the CC and never worry about the downhill going more than one mile over my setting and never going slower than allowed even on the steep uphill.

[strider]

Are you referring to how it behaves when you hit "resume"? It does speed up quite aggressively to get you back to your target speed but it's a lot smoother than an automatic ICE that tries to hold high gear and then kicks down a gear or two, engine screaming as it races back to your desired speed....

[slcasner]

Driving with cruise control on, but I desire to speed up a hair to get around someone. If I lightly touch the torque pedal, I seem to engage regen, because the car seems to slow. So I have to press past the "sweet spot" to get the car to speed up. This seems counter intuitive.

I'm not sure if regen is actually engaged, but I've noticed that the Roadster has the same behavior that I have disliked in ICE cars. While driving with the cruise control engaged, if you speed up by pressing the accelerator to pass another car, and then let off, the cruise control will let the speed decrease below the set speed before bringing the speed back up to the set speed. In an ICE, I interpret that as the cruise control backing all the way off for its actuation of the trottle because the car is going faster than it is supposed to. Then, when the driver's foot comes off the accelerator, it takes a while before the cruise control reacts to begin actuating the throttle again. In the Roadster, though, there should be no lag. It should be able to let the speed drop smoothly down to the set speed again without going below, just like it locks onto that speed very precisely.

My biggest gripe is that the minimum speed that the cruise control will hold is 30mph. I'd like to have that be 25mph for assistance in avoiding cops with radar guns on the residential streets around here.

For those who normally eschew cruise controls, when you are trying to maximize range on a freeway drive, it is a big help to keep from going faster than you intended, and to avoid acceleration/deceleration cycles. Those of us who had EV1s a decade ago learned this well.

[Doug_G]

I'm not sure if regen is actually engaged, but I've noticed that the Roadster has the same behavior that I have disliked in ICE cars. While driving with the cruise control engaged, if you speed up by pressing the accelerator to pass another car, and then let off, the cruise control will let the speed decrease below the set speed before bringing the speed back up to the set speed. In an ICE, I interpret that as the cruise control backing all the way off for its actuation of the trottle because the car is going faster than it is supposed to. Then, when the driver's foot comes off the accelerator, it takes a while before the cruise control reacts to begin actuating the throttle again. In the Roadster, though, there should be no lag. It should be able to let the speed drop smoothly down to the set speed again without going below, just like it locks onto that speed very precisely.

A cruise control is a type of servo control loop, which are usually implemented as PID loops (Proportional, Integral, Derivative). They can be implemented in analog or digital electronics, usually digital these days, but they work the same way -- it is all dictated by some very fundamental theory.

These loops have a bunch of constants that have to be tweaked to optimize the performance. If you set them wrong you can get completely unstable operation. The optimum settings are always something of a compromise between stability, the desired response rate, overshoot, undershoot, etc.

Proportional provides basic feedback. If the car goes too fast, it backs off the gas.

Derivative is for speed of response. Increasing it causes the servo loop to be more responsive to changes.

Integral is there to pull you in to the right speed. If you're just a hair off, the error integrates over time until it pulls you in. Of course if you overdo Integral it will cause you to overshoot. But you need it because otherwise you won't pull in to the right speed. Typically there is an integral limiter, so it doesn't go ape if the inputs change faster than the system can respond.

The problem you describe -- overshoot -- is fundamental to all servo control systems. You can improve on things by a suitable choice of constants, but there is no set of constants that will give you optimal behavior on all performance measures. You can analyze the system to death and do a lot of math and come up with the optimum solution, but in practice this is very hard and no more effective than an expert using his experience and judgement to optimize and balance the various constants. You can't eliminate overshoot, and trying to minimize it too much will result in poor performance in other areas. These things trade off against each other.

So unfortunately the overshoot you hate (me too, actually) isn't there because the designer was lazy. It's just the nature of servo loops. Maybe it can be made a little better, but there are fundamental theoretical and practical limitations to what is possible.

[S-2000 Roadster]

Are you referring to how it behaves when you hit "resume"?
No, I am forgiving of 'resume' behavior and have used it seldom. In the Roadster, I have not even figured out which button to push to get it to resume!


It's just the nature of servo loops. Maybe it can be made a little better, but there are fundamental theoretical and practical limitations to what is possible.
Doug, thanks for another excellent description (I also read your note on power and voltage regulation). I've only ever designed servo loops for very simple applications and dread the thought of having to get it right myself, especially considering the really bad implementations I've seen from amateurs. I assumed that it could be better, but based on your description above I can admit that it really wouldn't be possible to make it perfect for all conditions - at least not without combining some kind of learning neural network programming combined with the standard PID (if that is even possible).

[strider]

No, I am forgiving of 'resume' behavior and have used it seldom. In the Roadster, I have not even figured out which button to push to get it to resume!
The top half of the end of the stalk is a rocker switch. Push the dash-side end in to set. Push the person-side end to resume. The bottom half is cancel.

[S-2000 Roadster]

The top half of the end of the stalk is a rocker switch. Push the dash-side end in to set. Push the person-side end to resume. The bottom half is cancel.
Ah, 'R' is for Resume as well as Reduce?

I've managed to use the 'I' to increase speed if I set it too low, and 'R' to reduce speed if it settles 1 mph too high. Obviously, I must have found that 'I' works to Set the speed. I guess sometimes the system doesn't resume if the timing isn't just right. Then again, it could have been operator error, since one time I managed to hit the Off button instead of Resume, and the regen braking kicked in hard. Now that I think about it, I must have had Resume work at least once or twice, but I was remembering having more trouble with this cruise control implementation than any other.

Hmm, I wonder if the limited 8-bit nature of the accelerator input has anything to do with the lack of smoothness I notice.

[Doug_G]

I guess sometimes the system doesn't resume if the timing isn't just right.

Yeah, it's fussy on timing. You have to hold it down for just the right amount of time. I've also found it's harder to press the R side for some reason. Sometimes I don't bother with Resume because it's easier just to reset it.

[SByer]

The Roadster cruise control really sticks the speed you set it at. The part it doesn't do as well as I've seen (say, on my wife's Merc) is that it doesn't have a way of bumping the speed up and down by 1mph (or 1kph if you prefer the better units - as an engineer it makes me cringe that our government gave up on switching over when it was cheaper...).

The E wagon we have is great - you start the cruise control and then can bump it up or down to exactly the speed limit (yes, I know, there are variabilities in all sorts of things which means that's an estimate, but it totally satisfies the OCD in me.) It strikes me that the Roadster could easily have done that and, with the GPS feed into the VDS, been completely accurate.

But then, it's just cruise control, I don't ever really get a chance to use it. It's still far better than most cruise controls I've seen.

[vfx]

I have mastered the 1mph bump. Taping on the end of the stalk or the lower half gets the mph to go up or down respectively. It does take finesse and while learning you have to wait to see what your action does on the VDS readout.

[mpt]

A bit late but, the roadster cruise control is implemented in software. It runs in the next loop outside of the heart of the DMC so, after the code that generates the motor control and charge syncing loops, the next loops handle cruise, traction control, etc.

All I wish for it that it worked at 25mph round town.

[TEG]

...It runs in the next loop outside of the heart of the DMC so, after the code that generates the motor control and charge syncing loops, the next loops handle cruise, traction control, etc...

How do you know that? It sounds like a quote directly from one of the drive-train firmware engineers!

[mpt]

Digging... Digging... 2008 Drew Baglino, Senior Electrical Engineer, The Spin Stops Here | Blog | Tesla Motors (http://www.teslamotors.com/blog/spin-stops-here)

[TEG]

Digging... Digging... 2008 Drew Baglino, Senior Electrical Engineer, The Spin Stops Here | Blog | Tesla Motors (http://www.teslamotors.com/blog/spin-stops-here)


Awesome - thanks. I knew that quote sounded familiar. Good reference.

[ggr]

The part it doesn't do as well as I've seen (say, on my wife's Merc) is that it doesn't have a way of bumping the speed up and down by 1mph (or 1kph if you prefer the better units - as an engineer it makes me cringe that our government gave up on switching over when it was cheaper...).

Ummm, mine does (roadster 2.0). On the end of the blinker stalk there are the convex and concave buttons, and the larger flat "cancel" button. You press the convex one to engage cruise control at the current speed, or hold it to accelerate. But if you just "click it" (think like a mouse click), it raises the speed setting by 1mph. Similarly the concave button nearer the driver is the "resume/decelerate" button, and clicking it changes the set speed down by 1mph. This is pretty much the interface on all cruise controls I've ever seen. I don't remember whether it was documented in the manual, but I tried it and it works. (Yes it is... page 6-9 in my manual.)

I agree about the units. Even worse was England, which seemed to take most of the pain and then backtracked.

[mpt]

Just as a datapoint, the cruise switch is a GM part that dates back to the 1980's - here's a pic of an '87 Vauxhall Carlton GSI (ace car btw) - the behavior of the cruise control is all Tesla as it's software. For me the fact that it holds speed uphill and down wins it a reward. My Audi did the same by conspicuously putting on the brakes... not smooth!

2222

[ChargeIt!]

Ugh. The cruise control in the Roadster is dreadful. I'm a cruise control junkie and use it for ALL my driving in ALL my cars. The 1985 Corvette had the best non-adaptive cruise control ever made. The 2011 BMW 550i has the best adaptive cruise control ever made. (no bias here, I don't own either car).

The Roadster is horrible. Terrible ergonomics, stupid 30mph lower limit, forgets the "set speed" if the vehicle speed drops below 30mph, cancelling will throw you through the windshield, trying to overridge during passing is less than graceful, dropping back to the desired speed after passing is less than graceful, etc. I'd rate it a D-.
Did I mention "terrible ergonomics"? I better mention again just in case. Terrible ergonomics.

I just hope that it was some Lotus part and that the Model S will be much better.Agreed on the terrible ergonomics, but not all of your other points. And the "forgets below 30" is not correct; the limit, I think, is 2mph or a full stop. Resume works ok above that. Have yours checked. (And override is done with the go pedal.)

[S-2000 Roadster]

+1 for the terrible ergonomics, or should I say "-1" for Tesla Motors? I've hit the Off button when I wanted the two-way toggle, and that's a shocking mistake.

It does seem strange that mpt's photo of the GM part is in a right-hand-drive vehicle. What's the story there? It seems backwards that a British maker would source a crappy GM part, but I guess I shouldn't be surprised.

I've never had Resume work, and I certainly haven't dropped below 2 mph on the freeway! I'm going to have to test this some more.

You can still use the crappy behavior if you predict it well enough. I frequently use the Off button to slow down when I reach the freeway exit for home, and I time it right at the point where aggressive regeneration braking is what I want, rather than the horrible side-effect that it would normally be.

[Doug_G]

Resume requires holding down the button long enough, but not too long. It's hard to make it work.

I've also hit cancel when I wanted to slow down a little. Wham!

I actually like the way the cruise drives; I just don't like the control stalk.

[Nik]

What's the story there? It seems backwards that a British maker would source a crappy GM part, but I guess I shouldn't be surprised.


Vauxhall is a UK-only brand for GM - the same cars are generally sold as Opel in the rest of Europe.

[mpt]

I remember one of the tech blogs referring to the parts supply chain as being one of the hardest parts of the process. Mainly it was being able to secure supply in such low numbers. Part of the reason that small auto companies have to use parts-bin spares is not just the cost of design, engineering, functional testing, safety testing, fire testing, etc. etc. but that these auto factories will knock out 10,000 parts before you can reach for the off-switch. So, Lotus use the freely available GM part and hence Tesla follows suit. I recall that they had to almost beg to use the airbag in the wheel.


In the 90's Lotus had a fad for Peugeot interior parts; better I thought but probably more expensive.


I'm speculating that the choice of the big LCD display in the 'S' isn't just about style but about cost. Just think of all the controls, wires, cables, leavers, plastic parts, illumination sources and ducts that disappear when you just go for an LCD panel and an off-the-shelf CPU unit. And, it's your supply chain; not GM.


BTW: When I snapped that Cruise control stalk off I was very glad of its heritage; just $43 for a replacement Living with a Tesla Roadster (http://www.teslamotorsclub.com/showthread.php/3579-Living-with-a-Tesla-Roadster?p=40023&viewfull=1#post40023)


It seems backwards that a British maker would source a crappy GM part, but I guess I shouldn't be surprised.

...In the US this car is the Cadillac Cetera

Oh, and another thing; the prototype cars used the GM window switches; the early round ones; I loved those in the Cavaliers and Carltons, they had such a lovely 'click' when you pressed them, and they lit up. Stupid really but before the car shipped, in my mind I imagined clicking them; part of the whole experience. Crazy eh? Ooh, I wonder if I could retro-fit them...

[S-2000 Roadster]

Oh, and another thing; the prototype cars used the GM window switches; the early round ones; I loved those in the Cavaliers and Carltons, they had such a lovely 'click' when you pressed them, and they lit up. Stupid really but before the car shipped, in my mind I imagined clicking them; part of the whole experience. Crazy eh? Ooh, I wonder if I could retro-fit them...
Ooo, do it! ... then send photos. I wish I knew more about the Elise so I could help with the research (assuming that there are easy-fit options that could be moved over from Elise to Roadster).

My S2000 has window switches that click twice. Part way down will only operate the window motor while you're pressing (momentary); all the way down to the second click will actuate the automatic motor mode where it doesn't stop until the window is all the way down. This is not only ergonomic and intuitive, but it is 100% reliable. Lifting the switch is a third contact to bring the window up, but only so long as you're holding it - in that respect they're the same as the Roadster when raising the window.

The Roadster switches are fine, but I really hate the picky timing. Sometimes I'm too fast for the tap-to-auto-lower-to-the-bottom. Then I hold too long and get the momentary function. I think that the timing of the cruise control must be the same: too fast on the Resume and it doesn't Resume; too long and it Reduces speed instead of Resuming speed. It's like the main computer is reading all of these switches but the interrupt response is too slow to react to faster human inputs.

[ChargeIt!]

My comments were for Roadster 1.5. Anyone else with 2.0+ that have experienced the "forgets below 30" issue ?

[S-2000 Roadster]

My acura very easily keeps the exact speed, even down a steep hill. It will even downshift instead of riding the brakes all the way down.
I might be way off here, but I don't think that any cruise control system ever touches the brakes. At the most, you'll get engine compression braking (ICE) or regenerative braking (EV).

[TEG]

In the late 80's I built my own fuel injected engine for my drag car, and also built my own fuel injection car computer and programmed my own cruise control software (all in 6502 machine language no less)...

Ha! In the late 80s I built my own digital dashboard / car computer based on a Vic20 motherboard all in 6502 (6510) assembly as well!
But I think you got me beat on the fuel injection part. I didn't try to touch the actual drive-train interface... Just lighting and monitoring.

[Nik]

Ha! In the late 80s I built my own digital dashboard / car computer based on a Vic20 motherboard all in 6502 (6510) assembly as well!
But I think you got me beat on the fuel injection part. I didn't try to touch the actual drive-train interface... Just lighting and monitoring.

Hey. I had a VIC20 too. I didn't use it for anything to do with cars though, I was only 10...

Is the tangent complete?

[mnx]

My BMW applies the brakes on hills to maintain the set cruise speed. it's annoying... I try not to use cruise control in hilly areas for that reason.


I might be way off here, but I don't think that any cruise control system ever touches the brakes. At the most, you'll get engine compression braking (ICE) or regenerative braking (EV).

[Doug_G]

My comments were for Roadster 1.5. Anyone else with 2.0+ that have experienced the "forgets below 30" issue ?

No. The cruise will disengage but it won't forget the setpoint. I think a "rolling stop" is enough to keep it.

[Doug_G]

I might be way off here, but I don't think that any cruise control system ever touches the brakes.

The adaptive cruise control in my Infiniti G37 does.

[Lloyd]

Mercedes does as well.

[ChargeIt!]

More info.

@William3: Sounds as if we're both right, although it depends on your definitions :wink:

Again my comments are limited to Roadster 1.5. Below 30 the RESUME feature does NOT work, but that does not mean that the car forgot the setpoint. So ... in your 2.0+ ... try this: accelerate to 32mph; set cruise; after short while cancel cruise (with, for example, brake pedal). Decelerate below 30, try RESUME; it won't work. Accelerate to above 30, let's say even higher than setpoint, like 42. Try RESUME; it works (let "go" of GoPedal). The car will settle back to 32. (The same procedure and result with a higher setpoint, like 55, etc.)

:smile:

[hcsharp]

...
What I'm saying is that my 2010 Roadster will forget the set speed if the vehicle speed drops below 30mph.
My version 2.5 has this same behavior.

[markwj]

forgets the "set speed" if the vehicle speed drops below 30mph

I just tested with my 2.5 base roadster (delivered in spring 2011).

1] I went to 70kph and engaged cruise control. Everything good.
2] Dropped out of cruise control and decelerated to 60kph, resumed cruise control and car took over back to 70kph set speed.
3] Dropped out of cruise control and decelerated to 25kph. Car refused to resume cruise control. Accelerated back up to 60kph, resumed cruise control and car took over back to 70kph set speed.
4] Dropped out of cruise control and came to a complete stop at a traffic light. Accelerated back to 60kph, but car refused to resume cruise control.

The manual says:


Resuming a set speed

Press R to accelerate to the previously set speed. This function will not operate if the previously set speed is zero or if cruise control is already operating.

The set speed is reset to zero when:
1. The starter switch is turned to the OFF position.
2. Reverse gear is selected.
3. The hand brake is applied.
4. The vehicle speed is below 1.6 km/h (1 mph).

That seemed to match what I experienced.

If your car is resetting the set speed to zero in anything other than those four conditions, then it is broken (or the 2010/2011 manual and car I am driving is different than yours). Mine is UK spec.

That said, the cruise control in the roadster is far from ideal. I can't understand the logic for limiting it to >30mph(48kph) and the buttons are ridiculously hard to press.

[Doug_G]

markwj your experience matches mine (2.0 base Roadster).

Every car I've driven has a lower limit on the speed. Not sure why; perhaps there are safety concerns?

Agree that the buttons are ridiculously hard to press. Especially the slow/resume button.

[bolosky]

I tested my early-ish 2.0 non-Sport (#670) last night. It does the same thing as markwj: It won't resume when it's under 30mph, but it does remember the set point.

[markwj]

Thanks for the input guys. Hopefully a few more USA 2010 Roadster owners will chime in, then I'll have enough evidence to bring this up with the Tesla service department again.

I suggest you check your manual, then try to repeat what we did. Set point somewhere above 30mph. Slow down to below 30mph (but don't stop). Speed up to above 30mph. Then resume. If car doesn't resume, and the manual says it should, then your car is not performing to spec.