difference between visual geometry and collision geometry in the simulation

[jxtxwsbn]

Hi,

I noticed that there is a difference between visual geometry and collision geometry. The gripper can grasp the object even there is a gap between the finger and the object. What's the reason for this phenomenon? Thanks

[mbreyer]

This is a behavior of the physics engine. PyBullet recommends that you generate a convex decomposition of the visual geometry using v-hacd since computing collisions between convex objects is a lot more efficient. It can lead to some artifacts though when the convex approximation does not exactly match the visual mesh.