Nonlinear Model Predictive Control Schemes for Obstacle Avoidance

This work proposes single-layer nonlinear model predictive control schemes to solve the autonomous navigation problem while providing obstacle avoidance feature in cluttered environments with previously unknown obstacles. Considering model predictive control frameworks for set-point stabilization and set-point tracking, the penalty method of nonlinear programming is taken into account to enforce avoidance constraints without losing stability and feasibility guarantees. The set-point tracking schemes are shown to be more suitable for motion systems due to their enlarged domain of attraction with respect to the regulation formulations, making it feasible for any changing targets. Further, for the set-point tracking problem, the proposed schemes avoid the use of terminal regions, which, for nonlinear systems, might be cumbersome to compute. Thus, simple design schemes based on a relaxed terminal equality constraint and on a weighted terminal cost are considered. Finally, two case studies considering a differential mobile robot and a quadrotor unmanned aerial vehicle are provided to evaluate the set-point tracking formulations.