Connectivity maintenance in ad-hoc UAV networks for multi-robot missions

Abstract

Connectivity maintenance is an important problem in multi-robot systems, which requires designing control algorithms to maintain network connectivity at a desired level. This work implements connectivity control algorithms in the context of a specific multi-robot scenario. Additionally, this controller is modified to be ’battery aware’ (BA) and minimizes the risk of connectivity loss in case of agent failures. The proposed BA controller shows superior performance compared to classic control, especially during transitional phases. Multiple ex periments are conducted with different topologies to generalize the performance of the pro posed controller. This work also introduces Graph Neural Networks (GNNs) for imitation and reinforcement learning of the proposed controllers. The applicability of the proposed algorithms is demonstrated using Crazyflie UAVs in both simulated and real environments.

9