Decentralized cooperation of multiple UAS for multi-target surveillance under uncertainties

An interesting applications of Unmanned Aerial Systems is surveillance. Surveillance typically involves the tracking of one or several targets in an area. A key issue for this application is the autonomous decision-making to allocate UAS to targets and determine the actions to be performed by the UAS of the fleet. Since this optimal decision-making needs to deal with uncertainties, Partially Observable Markov Decision Processes (POMDPs) are proposed as models for the surveillance mission. The paper proposes a role-assignment method to alleviate the computational complexity of the application of POMDPs to multi-UAS surveillance. The method is evaluated by simulation and compared with other similar approaches. Furthermore, the system has been implemented in a testbed with real quadcopters to show some preliminary experiments of two UAS tracking two different targets.