Sparse Shooting at Adaptive Temporal Resolution for Time-Optimal Model Predictive Control

This contribution presents a novel approach for time-optimal model predictive control. The underlying optimal control problem rests upon an adaptive, local temporal discretization of the shooting grid and automatically determines optimal switching points in the control sequence. Furthermore, the grid size is adapted online in order to solve the control problem with a minimum number of control interventions. The approach offers significant advantages for bang-bang control tasks which exhibit few transitions between a discrete set of piece-wise constant control actions. Experiments and a comparative analysis on different nonlinear control tasks demonstrate the superiority of adaptive shooting grids w.r.t. state-of-the-art approaches in model predictive control.