# [ODE] Collision: number of intersection points

Thomas Thueer ode at thoemsen.ch
Sat Feb 12 14:35:12 MST 2005

Hi Megan

>Remember that ODE is working with perfect spheres, so if you imagine a
>non-penetrating collision between an infinite plane and a perfect
>sphere, you'll see it has exactly one point of contact, infinitely
>small.
>
>
OK, that's logical, but there are two things that confused me:
- Sometimes dCollide() returns more than one point of intersection
between the sphere and the ground, so I guess that was when the sphere
collided with 2 different triangles of my mesh-terrain.
- When I allow penetration of the bodies in ODE (CFM), I thought that
this means that ODE calculates the real penetration depth and
intersection of the bodies, which does not seem to be the case according

>
>If you're not getting enough traction, then why not ramp up the
>friction of contact joints generated between your wheels and the
>ground?  (make sure you're using the Friction Pyramid Approximation
>friction, which is not the default friction model)
>
>
I already increased the friction coefficient (even up to dInfinity)
using the Friction Pyramid, but it didn't really help. So I decreased
the slip on the contacts, but that caused ODE to find incorrect
solutions for the movement, I guess because the system was somehow
overconstrained (the middle wheels of a 6-wheeled rover were lifted off
the ground as soon as the front wheels touched a step obstacle, so the 4
wheels moved all vertically at the same time!). I have played a lot with
the contact joint parameters, but my problem is that I need a solution
that works fine for any rover-type vehicle because we are developing a
simulator and cannot ask the end user to find the parameters for every
vehicle. Do you have any advice?

One last question: in reality there is a big difference between
wheel-ground interaction of a sphere and a cylinder because of the
contact surface. How would you model this difference in ODE then without
setting different friction coefficients?

Anyway, thanks for your help so far.
Thomas