Heterogeneous multi-robot system mission planning with cooperative replenishment through data-driven rendezvous point selection

Abstract With advances in autonomy technology, the use of multi-robot systems is becoming increasingly viable and efficient. In particular, heterogeneous systems are effective in complicated missions that require various capabilities. When the missions are prolonged, certain robots such as quadcopter-typed unmanned aerial vehicles (UAVs) run out of energy and require replenishment. As some robots such as unmanned surface vessels (USVs) have large payload capacities for carrying supplies, the UAVs can receive the necessary resources from the USVs amid the missions. In this paper, we propose a mission planning framework that aims to minimize the total mission duration with consideration of the energy replenishment of robots and the heterogeneity of robots and tasks. The proposed framework consists of four components: task allocation, rendezvous point selection, task planning, and plan combination . For replenishment, the location of the rendezvous must be chosen, and we propose using a data-driven approach to predict the best rendezvous point. Then, the prediction result is used for computing candidate plans of each robot in a distributed fashion, and the best candidates are combined to compute the final plan. To validate the efficacy of the proposed method, simulated experiments of various problem configurations are performed and analyzed.