Distributed Multiple Model MPC for Target Tracking UAVs

In this paper, the idea of using teams of Unmanned Aerial Vehicles (UAVs) to track a ground vehicle and exploiting the benefits of multiple UAVs is considered. The design and testing of a Distributed Multiple Model MPC (DMMMPC) controller for tracking in formation flight is investigated. Using information from state estimation about which target model is performing best, the DMMMPC changes its target motion model accordingly to match the target. This MPC controller is first implemented for a single UAV, then tested in both a real-time simulation environment and indoor flight. The MPC is then expanded to the multi-UAV scenario, which is tested in the same real-time simulation environment, demonstrating effective target tracking in formation flight for the team of UAVs. Lastly, the strength of the distributed topology is shown by exposing specific agents in the formation to measurement occlusions and observing the degradation of the tracking performance of the occluded UAVs.