Systematic design of predictive control for autonomous surface vehicles in path following with obstacle avoidance

Abstract

A systematic procedure to design the spatial constraints for a predictive control strategy focused on path following of autonomous surface vehicles is proposed to avoid obstacles. Firstly, a systematic design algorithm is devised to generate a set of positional constraints to be imposed on the predictive controller from the map of the aquatic body and the obstacles. Secondly, a recently presented non-linear model predictive control-based guidance strategy for path following is extended to consider output constraints, such as those generated by the previous design algorithm. This controller has been shown to overcome drawbacks of other line-of-sight-based guidance laws and to enable the application of predictive strategies to the low-level controller through the computation of future references for the forward velocity and heading angle. Finally, a practical predictive strategy has been tailored to this problem to generate a simplified version of the proposed predictive controller by linearizing the model along the trajectory. Moreover, the positional constraints describing the obstacles and the aquatic body are adapted so that only linear constraints are imposed in the optimization, giving rise to a much faster implementation. The effectiveness of both proposed control strategies in navigation and obstacle avoidance in a complex and realistic scenario is illustrated in simulation, where their path-following performance and computational cost are compared.