Data-driven approach to ensuring fault tolerance and efficiency of swarm systems

Swarms of drones are increasingly deployed to perform a variety of monitoring tasks. However, despite a growing popularity of swarm-based monitoring solutions, there is still a lack of approaches that allow the designers to ensure that a swarm configuration maximizes coverage of the monitored area and delivers the desired quality of the monitored data. A drone might fail or run out of energy and hence leave some blind spots in the monitored area. Moreover, an inefficient configuration resulting in the energy-greedy communication can quickly deplete the batteries of the drones and force a premature mission termination. We propose a novel data-driven integrated solution that combines learning and optimization over the streams of monitored data to ensure fault tolerance and efficiency of the swarm navigation. The benchmarking results demonstrate that the proposed approach allows us to significantly improve the coverage and energy efficiency characteristics.