Extensive analysis of Linear Complementarity Problem (LCP) solver performance on randomly generated rigid body contact problems

摘要

The Linear Complementarity Problem (LCP) is a key problem in robot dynamics, optimization, and simulation. Common experience with dynamic robotic simulations suggests that the numerical robustness of the LCP solver often determines simulation usability: if the solver fails to find a solution or finds a solution with significant residual error, interpenetration can result, the simulation can gain energy, or both. This paper undertakes the first comprehensive evaluation of LCP solvers across the space of multi-rigid body contact problems. We evaluate the performance of these solvers along the dimensions of solubility, solution quality, and running time.