Centralized multi-robot trajectory planning via multi-stage grouping.

Abstract

In this study, we present an approach to effi ciently generate deadlock-free optimal trajec- tories for scale multi-robot systems with intensive space sharing setting or space-critical environments. The proposed solutionfirst individually plans a discrete path through a graph-based planner, and then refi nes this solution into smooth trajectories using coupled nonlinear optimization. For the coupled optimization, the robotfleet is split into groups that will be separately coupled in the optimization problem, where the current workfirst proposes a heuristic to dynamically group thefleet over the trajectory steps, and second to have multiple stages of optimization, one stage for each diff erent group set. Verifi cation is conducted through simulation and experiments. Compared to decoupled trajectory op- timization, the proposed approach reduces computation time on the defi ned test set up to 48.6%, and maintain the success rate of thefleet. The study results in a scalable solution that increases the computational effi ciency for space-critical multi-robot scenarios.