An optimization-based algorithm for Coulomb's frictional contact

Abstract

The main goal of this paper is to propose a stable algorithm to compute friction forces governed by Coulomb's law in the course of the simulation of a nonsmooth Lagrangian dynamical system. The problem appears in computational mechanics, to simulate the dynamics of granular materials, robots, etc . Using a classical impulse-velocity formulation of Coulomb's law to model friction, and a semi-implicit time discretization scheme, we get a set of linear, non-linear and complementarity equations which has to be solved at each timestep. Two mutually dual parametric convex optimization problems coupled with a fixed point equation appear naturally. By solving (one of) these optimization problems iteratively within a damped nonsmooth-Newton algorithm, it is possible to decrease some infeasibility criterion and hopefully converge to a solution of the system. Numerical results are provided, which show that the number of iterations needed by the algorithm is very small in general and that the method is stable.