Real-time cross-domain monitoring of multi-UAV-multi-USV systems via efficient block sparse Bayesian learning

Despite the tremendous progress in coordinating multi-unmanned surface vehicle (USV) fleets, persistent monitoring remains a dilemma because USVs cannot share data with external monitors. Practical deployments further impose real-time constraints and limited onboard calculation capability, necessitating low-complexity algorithms. This study proposes a multi-UAV fleet-based monitoring scheme. Therein, UAVs are assigned to pairwise USV-UAV matching to observe relative positionreal time. An efficient block sparse Bayesian learning algorithm (EBSBL) is then developed to identify the coordinated dynamics of USVs, with theoretically guaranteed feasibility. In addition, the unscented Kalman filter (UKF) is employed to facilitate multi-UAV coordinated monitoring with real-time prediction and USV trajectory estimation. The effectiveness and superiority of the proposed method are demonstrated by both numerical simulations and real-lake based multi-UAV-multi-USV platform experiments.