Convex Quasi-Dynamic Simulation of Rigid Point Clouds with Torsional Friction
Keywords: simulation, optimization, point cloud, physics, quasistatic, quasidynamic, quadratic programming
TL;DR: A fast and robust method for quasi-dynamic simulation of rigid point clouds with torsional friction at each of the points for more realistic behavior.
Abstract: Many common manipulation tasks move slowly enough that velocities and Coriolis forces do not contribute meaningfully to the dynamics of the robot. As a result, quasi-dynamic simulation of these systems, in which all forces are assumed to be in equilibrium, can produce physically accurate results for use in motion-planning and control. Recent work has combined the quasi-dynamic model with a relaxed formulation of Coulomb friction, where the resulting dynamics are the solution to a convex quadratic program. We extend this recent work by directly manipulating rigid point clouds, without the need for meshing or decomposition into convex primitives. We also introduce a novel torsional friction model to mimic the frictional behaviors of the sorts of patch contacts that exist on real systems. These ideas are demonstrated in a grasping example, where a dense point cloud is manipulated with and without torsional friction, clearly showing the utility of the torsional friction model.
Submission Number: 4
Loading