Centralized Versus Distributed Nonlinear Model Predictive Control for Online Robot Fleet Trajectory Planning.

In this paper, we formulate and evaluate a centralized vs. a distributed approach for online trajectory generation for a fleet of mobile robots in the presence of static and dynamic obstacles. Due to dynamic obstacles, the trajectories need to be updated online and this is formulated as a nonlinear model predictive control problem. We show that both centralized and distributed MPC solvers manage to generate smooth collision-free trajectories. The distributed approach is shown to scale to many robots very well. In contrast, the computational cost of the centralized approach increases with the number of robots. However, the trajectories generated by the distributed control approach have larger deviations than those generated by the centralized approach. The experiments suggest that the centralized method should be chosen with sufficient computation resource while the distributed approach is a viable alternative when the number of robots is considerable.