Heading and position receding horizon control for trajectory generation

This paper presents a comparative study of two formulations of model predictive control with receding horizons for the cooperative control of a team of unmanned aerial vehicles. In the first formulation, the vehicle trajectories are solved dynamically as sequences of vehicle headings over prediction horizons and executed over shorter action horizons. This formulation takes advantage of an implementation of collision avoidance-based on vehicle heading constraints. In the second formulation, the vehicle trajectories are solved as sequences of vehicle positions, rather than vehicle headings. This formulation handles collision avoidance with vehicle position constraints. An efficient branch-and-bound algorithm is proposed to support the mixed integer constraints, and a collision avoidance solution based on heading constraints is evaluated. This paper shows that both receding horizon formulations produce exactly the same vehicle trajectories when they are used without collision avoidance constraint. It is shown however that heading receding horizon control requires less computing power than position receding horizon control whether in situations of collision avoidance or not.