[ODE] Yet Another Car Simulation With Problems

Gabriel Gambetta ggambett at artech.com.uy
Wed Aug 18 10:47:28 MST 2004


Hi ODE guys,

I'm not an ODE expert in any way, and I didn't have many physics courses during my degree, so be easy on me...

After reading lots of documentation, examples and demos, I managed to have the usual four-wheeled car made of a box mass (using many box geoms to approximate the shape of the chassis) and 4 spherical wheels connected to the chassis by hinge2 joints, running over a heightmap-based terrain and objects such as ramps.

The simulation doesn't explode and the motion of the car is realistic enough for my purposes (simple, arcade-ish racing game).

However, I can't get the steering right. At high speeds (say 200 km/h), I turn right, the car goes to the right for a while, but when the steering wheels return to the neutral position, the car turns left very quickly, often doing a full 360º.

I've been playing with many of the parameters but I can't get a combination that feels good. I do have several ideas about how to improve this, and I'd really appreciate to hear your comments about them :

1) Chassis mass = 1000 kg, wheel mass = 20 kg; is there any inherent problem with these values if I keep everything else with coherent scale (the car size is in meters, for example)?

2) Damping - I've seen damping used several times; is it to stabilize the simulation or to simulate aerodynamic drag?

3) Progressive steering? Right now I'm using "digital" steering, that is, if the user presses RIGHT I tell the joints to go full right, not setting the angle but the speed. Do you think accumulating a "target angle" instead of using the full angle as target from the beginning could help?

4) Collision model - I tried using a slip model for the wheels but I couldn't find a good set of { surface_mu, tyre_mu, tyre_slip } that didn't either make the terrain as an ice surface or made the car tip over easily (which it didn't before since the CoG is at the same height as the wheels).

Any comments on these points and any other comments are very welcome!

Thanks,
	--Gabriel




More information about the ODE mailing list