Car physics: Computer assists

It's time for me to implement computer assists in our simulation. This includes the following: -ABS -Traction Control -Automatic Yaw Control--Stability Management I'm wondering if any of you out know the details of how these really work on real cars. Here's what I already know, and don't know: ABS: I know systems detect wheel slip on individual tires, and then start pulsing the brakes. What I don't know is exactly how wheel slip is detected, and how the brakes are pulsed. From an implementation viewpoint, when does car/wheel enter ABS mode, and how is brake force/torque pulsed when in ABS mode? Traction Control: This is actually probably easier to implement than ABS, because as soon as wheel slip is detected for a powered wheel, I can just cut engine power, or apply brakes to those wheels, or both. That's what I've read actually happens. Anybody know otherwise? Yaw Control--Stability Management: This is the big question mark. So far, I've read that based on many input conditions (steer angle, speed, lateral gees, etc.), the computer can apply brakes to inside/outside wheels to control the yaw of the vehicle to match the desired yaw. It can also adjust drive torque to inside/outside wheels through the differential to achieve the desired yaw. Now, what I've read so far has been very vague to say the least. Does anyone know the details of how this works? Thanks in advance for your help! -Dave Gierok

ABS doesn't detect wheel slip but actually wheel lock up. So it should be about as easy as detecting wheel spin for traction control. Of course I'm no programmer either. Any way on my car there are what's known as an exciter ring on each wheel hub or axle. These rings have a ribbed OD. There is a small pick up that senses these ribs. Now how the computer uses this info I'm really not sure. MadDAWG

Actually ABS systems do detect wheel slip. Maximum braking effectiveness is achieved at roughly 20% wheel slip. The computer monitors each wheel speed and if any wheel drops below the speed of the other wheels, the ABS system will start to retract an accumulator piston on that particular wheel to reduce the amount of pressure being applied to the effected brake system. It constantly monitors the wheel speed, and if it detects the wheel back up to speed, it releases the accumulator, applying more pressure to the effected brake system. It is these retractions and applications which feel like pulses in the brake system. The system will also activate the antilock function if all 4 wheels have a reduction in speed at a rate greater than what the controller is programmed to allow. This is how all 4 wheels are kept from locking in low traction situations like ice and such. The systems which I have been trained on work in this fashion, and each wheel effectively has a separate brake system, so that each wheel is able to maintain a slip of no more than 20%. The systems will try and maintain a 20% slip though, as that is what has been determined to be the most effective in terms of control and braking distance. Traction control can work two ways, with most systems being one or the other, and some even using both. Some of this technology may have changed, as I have been out of the business for about 6 years now, but that is what was being used then. Glen Pittman "MadDAWG" <[email hidden]> wrote in message news:[email hidden]... > ABS doesn't detect wheel slip but actually wheel lock up. So it should be > about as easy as detecting wheel spin for traction control. Of course I'mno > programmer either. > > Any way on my car there are what's known as an exciter ring on each wheel > hub or axle. These rings have a ribbed OD. There is a small pick up that > senses these ribs. Now how the computer uses this info I'm really notsure. > > MadDAWG > >

While its obvious you know more about this than I do, I don't see how an ABS system could measure wheel slip. Now this could be as simple as your definition of slip differing from mine. I would call slip an increase in wheel speed. A good example of this would be loosing traction due to an increase in power. Now if your calling slip as a locked tire sliding on the road surface then I can by it. Now back to how my 95 Mustang works there is basically a speed sensor on each wheel. If it was seeing an increase of speed due to slip (my style) then there would be no need for the Anti-LOCK to kick in because the wheel is not locked or close to it. So back to the question of game programming I would think it would be similar too, but opposite of traction control. I'm not questioning if your right or wrong, just trying to make sence of it is all. Thanks MadDAWG

On 3 Jan 2003, Dave Gierok wrote:

Quoted text (8 lines)

> It's time for me to implement computer assists in our simulation. > This includes the following: > -ABS > -Traction Control > -Automatic Yaw Control--Stability Management > > I'm wondering if any of you out know the details of how these really > work on real cars. Here's what I already know, and don't know:

<snip> About 6 or 7 years ago we met one of the field engineers responsible for tuning an anti-lock brake system for a particular car model. This was a 6-week process with several engineers assigned to each car to be tuned. We did not meet the programmers that actually wrote the ABS control program. Tuning this system was pretty hairy, he used a laptop to adjust parameters (integer math in the micro-controller at that time) while the car was tested on all sorts of surfaces. The controller program had ~1000 adjustable parameters, of which about 200 were used in the tuning process, the other 800 were generally left at the original "default" settings. It seemed that much of the ABS controller program was dedicated to producing a reliable estimate of vehicle speed when various wheels are all slipping. Another large part of the code was full of logic that branched to different control algorithms after it determined the type of road surface that the tire was running on. While the intent is to hold wheels at some ideal slip ratio for max braking, the reality is that wheel speeds are jumping all over the place during an ABS stop. I watched from the side of the road when a Corvette did some braking tests and each wheel came to a visible stop several times (and then immediately spun back up to speed) in the course of a 60-0 mph stop. So my conclusion is that you don't want to re-create an actual anti-lock program. These are all trade secrets, no manufacturer that I know has exposed ABS algorithms or source code. Your sim must know the actual velocity over the road, which should make it easy to keep wheels from locking up. Your tire model better include friction circle effects<grin>. -- Doug Milliken www.millikenresearch.com

Slip in braking would be a decrease in wheel speed, not an increase. "MadDAWG" <[email hidden]> wrote in message news:[email hidden]... > While its obvious you know more about this than I do, I don't see how anABS > system could measure wheel slip. Now this could be as simple as your > definition of slip differing from mine. I would call slip an increase in > wheel speed. A good example of this would be loosing traction due to an > increase in power. Now if your calling slip as a locked tire sliding onthe > road surface then I can by it. > > Now back to how my 95 Mustang works there is basically a speed sensor on > each wheel. If it was seeing an increase of speed due to slip (my style) > then there would be no need for the Anti-LOCK to kick in because the wheel > is not locked or close to it. So back to the question of game programmingI > would think it would be similar too, but opposite of traction control. > > I'm not questioning if your right or wrong, just trying to make sence ofit

Quoted text (7 lines)

> is all. > > Thanks > MadDAWG > > >

Quoted text (7 lines)

> On 3 Jan 2003, Dave Gierok wrote: > > > It's time for me to implement computer assists in our simulation. > > This includes the following: > > -ABS > > -Traction Control > > -Automatic Yaw Control--Stability Management

It looks like my posts aren't showing up so I'll repost this one. Ditto everything said about ABS. For traction control, the computer uses the same sensors and can limit wheel spin by closing the throttle, retarding the timing and/or applying (pulsing) the brakes. Stability management works with speed sensor, steering angle sensor (mounted on the steering shaft) and "G" force or yaw sensors. If the steering angle is more than expected for the speed and cornering force, the computer thinks the car is pushing and can apply the brake to the inside (of the turn) rear wheel thereby causing a brake "pull". Hope this helps and if I'm wrong on anything here, feel free to correct it as I'm still learning this stuff. Bill Berry

"Haqsau" <[email hidden]> wrote in message news:av74h6$cibtb$[email hidden]... > Slip in braking would be a decrease in wheel speed, not an increase. > Thats pretty much what I was figuring, but it seems odd to call it that. Oh well guess thats why they didn't ask me. MAdDAWG

Thanks Glen. Everything I've read so far agrees with what you've said here. I guess the biggest question is how the computer keeps track of the car's speed--if the car is braking, all wheels will probably be slower than the car's speed. Doug's post makes me think that the algorithm used is more complicated than I should try to imitate. Since this is just a game, I can cheat since I already know current speed and slip on each tire. Let me know if you have any more insight on the algorithm specifics of ABS. Thanks, Dave

Quoted text (8 lines)

> So my conclusion is that you don't want to re-create an actual anti-lock > program. These are all trade secrets, no manufacturer that I know has > exposed ABS algorithms or source code. > > > Your sim must know the actual velocity over the road, which should make it > easy to keep wheels from locking up. Your tire model better include > friction circle effects<grin>.

Of course, of course. I am trying to use your book as a bible of sorts :). Yikes, ~1000 parameters. I believe I'll 'cheat' a little bit and use our known vehicle speed, and slip values for each wheel. Actually, I already tried this and got very nice results (for ABS and Traction Control). I'll probably leave it at that. Yaw/Stability control is still difficult though. I plan on braking outside wheels in an oversteer situation, and braking inside wheels in an understeer situation. But detecting oversteer & understeer is trickier than I originally thought. Any ideas? Keep in mind we can 'cheet' again and use all current physical attributes that are already calculated in the simulation. Thanks, Dave

Quoted text (10 lines)

> Stability management works with speed sensor, steering angle > sensor (mounted on the steering shaft) and "G" force or yaw sensors. > If the steering angle is more than expected for the speed and > cornering force, the computer thinks the car is pushing and can apply > the brake to the inside (of the turn) rear wheel thereby causing a > brake "pull". > Hope this helps and if I'm wrong on anything here, feel free to > correct it as I'm still learning this stuff. > > Bill Berry

Thanks for the input. Any idea on how to detect understeer/oversteer situations such that we know when to brake inside/outside wheels? I can't immediately see how. Hmmm, anybody else? -Dave Gierok

You are doing a sim right? So you know the slip angles of the wheels? In the bicycle handling model (yes I go waaaay back), the understeer number (forget the exact name) is defined as the front slip angle minus the rear slip angle. If it's positive you have understeer, if it's negative you have oversteer. I presume with four wheels you could average them left to right and accomplish the same thing. Braking the outside wheels for oversteer? I suspect that is not what you want to do. You don't really want to apply a yaw moment to the car IMO, that would confuse the driver. Plus braking the outside rear could actually make things worse. What you want is to reduce the rear slip angle and/or increase the front. The trick is how do you do that using only the brakes, without causing so much forward weight transfer that it actually increases the oversteer? Don't know for sure, but I think if you just brake the front wheels a bit and leave the rears alone it might work. Or maybe just the outside front. Just thinking out loud really. ;o) "Dave Gierok" <[email hidden]> wrote in message news:3e19ce6c$[email hidden]... > > Of course, of course. I am trying to use your book as a bible of sorts:).

Quoted text (7 lines)

> > Yikes, ~1000 parameters. I believe I'll 'cheat' a little bit and use our > known vehicle speed, and slip values for each wheel. Actually, I already > tried this and got very nice results (for ABS and Traction Control). I'll > probably leave it at that. > > Yaw/Stability control is still difficult though. I plan on braking

outside > wheels in an oversteer situation, and braking inside wheels in anundersteer

Quoted text (9 lines)

> situation. But detecting oversteer & understeer is trickier than I > originally thought. Any ideas? Keep in mind we can 'cheet' again and use > all current physical attributes that are already calculated in the > simulation. > > Thanks, > Dave > >

On Mon, 6 Jan 2003 10:47:53 -0800, "Dave Gierok" <[email hidden]> wrote:

Quoted text (13 lines)

>> Stability management works with speed sensor, steering angle >> sensor (mounted on the steering shaft) and "G" force or yaw sensors. >> If the steering angle is more than expected for the speed and >> cornering force, the computer thinks the car is pushing and can apply >> the brake to the inside (of the turn) rear wheel thereby causing a >> brake "pull". >> Hope this helps and if I'm wrong on anything here, feel free to >> correct it as I'm still learning this stuff. >> >> Bill Berry > >Thanks for the input. Any idea on how to detect understeer/oversteer >situations such that we know when to brake inside/outside wheels?

Describe in 'AI Game Programming wisdom', chapter 9 something. In short: you take the wheel velocity versus car velocity at the wheel; take the signs of these. Then there are 4 situations, if you average the front and rear velocity differences; ++ -+ +- and --. These 4 represent understeer, oversteer and spin (I think understeer is there in 2 signs states). Don't have the actual source here, but I did an understeer/oversteer indicator (with the amount) for Racer (an unreleased version still). Gives you an indication of the amount of u/o steer as well. Quite nice and not often seen. :) Ruud van Gaal Free car sim: http://www.racer.nl/ Pencil art : http://www.marketgraph.nl/gallery/

On Mon, 6 Jan 2003 10:47:53 -0800, "Dave Gierok" <[email hidden]> wrote: ... >Thanks for the input. Any idea on how to detect understeer/oversteer >situations such that we know when to brake inside/outside wheels? BTW In the book 'AI Game Programming Wisdom' it uses the understeer/oversteer calculations to do countersteering for AI (as the amount of steer is a worthy value as well as its sign). Lovely chapter on racing AI, although a bit short (but worth buying alone for it). Ruud van Gaal Free car sim: http://www.racer.nl/ Pencil art : http://www.marketgraph.nl/gallery/