A coordinate descent approach to optimal tracking time allocation in point-to-point ILC

Iterative learning control (ILC) is designed for applications involving multiple executions of the same task. Existing work has applied ILC to point-to-point motion tasks, but has not fully exploited its design freedom to optimize performance criteria other than the tracking accuracy. This paper extends the task description of the point-to-point ILC framework for discrete-time systems by considering the tracking time instants of desired positions as changing variables (i.e. the temporal location of each position can vary). This extension allows the optimization of an additional performance index while maintaining the tracking accuracy. This optimization problem is solved using a two stage design framework, and an iterative algorithm consisting of a norm optimal ILC update and a coordinate descent approach is then derived to minimize an additional performance index, e.g. control effort, for the point-to-point motion tasks. This algorithm is tested on a gantry robot to verify its effectiveness in the presence of model uncertainty and disturbances