Poor physics performance with low mass/inertia

Looks like we roughly need a min 0.1kg mass and a 0.01kgm inertia at any axis/joint for even the base gc controller to be stable.

This is a bit weird (shouldn't be a problem) because when we project inertia into gc, the later links inertias are added into earlier links' entries in the inertia matrix. So we shouldn't really have problems.

Perhaps it's an issue with incorrect computation of centrifugal/coriolis forces. Should check this somehow. Maybe computing \dot {M(q) } is problematic.. Dunno.

Comments (0)