In this paper, we consider two approaches to iterative learning control for nonsmooth dynamical systems. If the system is passive, a proportional update law may be applied. For more general dynamical systems, we consider an iterative update algorithm based on a quadratic optimal control problem. The passivity approach does not require model information but is restrictive. The optimal control approach is general, but requires model information for its implementation. A mechanical system with dry

more »

... friction is used as a simulation example for comparison between the two approaches. Convergence is observed in both cases, but much faster convergence time is achieved with the optimal control approach.