15 replies to this topic

[ivanalesi]

Reading Nick Fry's interview yesterday, I got interested by this. I've never heard of this and I don't know what Fry calls driver in the loop simulator. Is it something like Williams' simulator or is it simulating g-forces on the driver? If it is the later one, then how do they work? If it's just like Williams' one, sorry for asking stupid question

[Ross Stonefeld]

Your sig is my thought.

[exFSAE]

Driver in the loop means its like driving a video game. Doesn't mean they simulate g-forces, and I doubt they do.

This is as opposed to a lap sim package which has a computer driver model that simulates a lap. Hard to get those right.

[ivanalesi]

Yeah, I googled a bit. It's like Williams' we've seen on TV. I always thought that vibration, heat, g-forces and even pain from the seat are needed to make a great simulator, yet they do just this projector screen instead...
Here's a press release from RBR's simulator:
http://www.qinetiq.c..._by_cueSim.html

Btw, some years ago, I had a device which I had to put on my head as sun-glasses and I could be in the game and nowhere else. I mean, you're driving the car at home, but you only see the cockpit and you don't see your desk or chair, it was funny and I think this technology was better than these huge screens:) But apparently it wasn't good for my eyes, because my eyes' focus was on my nose With the fast evolution in software and hardware, I guess in some years time we could buy one such simulator and beat Hammy ;)

[Ross Stonefeld]

Simulator pods can do pitch, roll, etc but you won't get gforce. I suppose it's an 'okay' substitute for feel but really what you want to have is top notch forcefeedback in the wheel.

[Greg Locock]

There aren't many 6 dof driving simulators around, Toyota has one. Here's another http://www.nads-sc.u.../facilities.htm

Slightly off topic, we've just installed a hexapod K&C rig, remarkably similar in concept to the one on this page. http://www.moog.com/...841/story_1923/

There's a lot of interest in *in the loop simulators, where * is the driver,and/ or the steering rack, or the tire, or say the shock absorber. So as you can see, Ross' comment about feedback through the handwheel is spot on. Of course we've done it the other way round and are using EPAS to degrade real world steering feel to Playstation steering.

To get real time response people tend to run simple physics based models rather than full blown MBS, so an increasing part of my job is to export ADAMS models as Vedyna models, then run the same thing through the same manouevres and check that the behaviour is similar (and that both are similar to the real world performance of course). A Vedyna model doesn't bother with suspension arms etc, it uses maps of camber vs wheel travel and so on, very like Lapsim. That is then fast enough to run in real time on a sensible PC.

You can modify an ADAMS model to run in real time on a network of very powerful computers (one for the car and one for each tire?), -I was told that one racing series uses real time modelling of the car during a race to see if/how the tires are going off, what setup changes to make, etc. However by the time you've stripped that much complexity out I think you'd be back to Vedyna levels of accuracy, so why not do it the easy way?

As a rough estimate 75% of the code is tires, 12% steering system, 13% is everything else, and frankly the tire models, steering rack models and shock absorbers are the weak links.

[zac510]

It's a funny wording though. How could the driver not be in the loop?

I understand the driver's input could be used 'open loop' in the simulator but that wouldn't seem to me to be a very good test when an engineer is changing the vehicle variables?

[Greg Locock]

Most of my tests are open loop. The classiest thing there is to use the actual steering inputs and speed trace from the correlation runs and then check that the model performs the same as the vehicle in every respect.

We also have path-following closed loop controllers that apply a feedback loop to the inputs to the model, and attempt to follow a defined course and speed profile.

A lot of our real world work is done with a steering robot to reduce 'noise'.

[zac510]

I see, following a set path makes a bit more sense

[ivanalesi]

Thanks Greg, you've got fun job:) I never thought it would be so hard to simulate the force feedback in the wheel. If you've worked with drivers, do you know if they really gain speed from such testing? Because replies from the F1 folk varies, from source to source, one driver says at one magazine it can help with setup, in another that it's just for learning the track direction...

[Greg Locock]

I don't do much work on racing cars, there are many people on this forum who can give a much better insight than me into whether playing on a playstation (or its big brother) helps the driver with lap times.

So far as car setup goes I know that in Peter Wright's book Ferrari claimed to be able to get pretty close to the final setup for a circuit just by modelling.

On production cars we can usually get close with spring rates and sta bars, not very good on jounce bumpers, not much help at all on dampers (other than just saying twice as much rebound as jounce, and 30% critical), and pretty good on boost curves and T bar. Black round things are our biggest problem.

[kangaloosh]

We do human/hardware/software in-the-loop simulation predominantly in the high-budget race team environment you describe.

It's not just about the driver, more that the driver becomes a component of the driver-in-the-loop simulator that (in some circumstances) is better than the driver-model, or robot, in the dynamic lapsim equivalent. If you model proposed engineering developments, your robot model will follow the track based on the rules you program, the real driver will take advantage of the new car in ways your model never dreamt of, finding new lines or grip that take advantage of the new car - for example, lets say you modelled a new control strategy for your diff - the real driver may find new ways to yaw the car into slow turns with the diff, that a model following peak combined slip ratios / slip angles / peak power will never benefit from.

I'm not saying that there aren't benefits for driver training, I think there are, I just don't think that is the major justification for the type of sim you posted about (cf Honda's Nick Fry quote in post 1 of this thread).

Worth bearing in mind that a driver-model or robot will always give you more consistency, (though top drivers can be fairly consistent), so in some circumstances driver-in-the-loop may not offer any advantages.

Given the current economic climate I think the cost benefits of this approach will become more important. I think testing in a warm and dry simulator, where the track is never rained off (unless you want it be I suppose), is probably a fair bit cheaper than dragging a car and associated engineers off to a test track? Taking F1 as an example - where next year they get KERS and adjustable front wing flaps, the control strategies for both of which are managed by the driver - I would be surprised if it were cheaper to manufacture those and test them in real life than it is to model them and test them virtually??

Interested to read Greg's comments, I have no experience of that level of hardware-in-the-loop sophistication yet, but we do run the real ECU as hardware-in-the-loop, so the driver can experience the control strategies programmed into the r/l hardware. You can mix this up with software-in-the-loop simulation (generally compiled Simulink) of new or proposed developments.

The best bit of all, that I've left 'til last, is that I get to drive the cars too (virtually of course )!!!

Chris

[lateralforce]

I've tried a 6DOF simulator. While the crashes are fantastic (since you do get tossed around), long high G corners are not felt because ultimately the actuators are limited to its travel limit in sustaining a low frequency G force. So does sustained straightline braking, and acceleration as well.

[Greg Locock]

I wondered about that. Basically those simulators are cheating, you can't reproduce all the accelerations just by tilting the body. That's not to say this couldn't be good enough, but it certainly isn't right.

[ivanalesi]

It depends on how important are these sensations to the driver, that's what I don't know. If Nico, Lewis or any other F1 driver can improve his driving from such simulator and be a better driver, than I guess they are not important, but F1 teams are so secretive and so conscious of their image, that they don't let the public know much about this technology.

[exFSAE]

The purpose of most of their simulation likely isn't to improve the driver. It's to improve the chassis (though driver training is always a plus). They need the driver in the loop because otherwise its very difficult to get a computation driver model that acts like you want, at the limit, and can adjust for different setups, racing lines, etc.