Numercial issue using HingeJointWithBacklash.proto

Dear RoboCupper,

There is an issue in the HingJointWithBacklash.proto:

During walking or kicking, Sweaty disappears sometimes unpredictable.
This phenomena can be reproduced, in case we lift Sweaty slightly and let it fall down:
http://www.hochberg.de/RoboCup/SweatyHingeJointWithBacklashOrig.gif

There is a workaround:
In case we increase the diagonals of the inertia matrix of the joint from 4e-9 to 4e-8, Sweaty never disappears.

The program ist stable, even if we drop Sweaty from an unreasonable height:

Probably other teams also have the problem of the sudden, unpredictable disappearance of the robot. In this case, we suggest a general modification of the proto.

We will upload our container using HingeJointWithBacklashHSO.proto.

Uli for Sweaty-Team

Original Proto:
http://www.hochberg.de/RoboCup/SweatyHingeJointWithBacklashOrig.jpg

Modified Proto:
http://www.hochberg.de/RoboCup/SweatyHingeJointWithBacklashModifiedInertia.jpg

Yes, this can be explained by the fact that the physics engine of Webots (ODE) shows instabilities when connected bodies have masses differing with several orders of magnitude, like 1 to 1000 or so. I guess the current Hinge2WithBacklash works well on lightweight robots like the Darwin-OP, but may cause the instabilities you observed on heavier robots. I believe it’s safer and also more realistic to increase the mass of the intermediate body included in the Hinge2WithBacklash proto as you suggested. So, we are going to apply this in the Hinge2WithBacklash proto, so that you can use it instead of the modified version you propose.

Thank you for investigating this problem and proposing a nice solution!