A novel decision-estimation methodology for a team of agents cooperating under communication imperfections is presented. The scenario of interest is that of a group of uninhabited aerial vehicles (UAVs) cooperatively performing, under communication delays, multiple tasks on multiple ground targets. In the proposed architecture, each UAV in the group runs an identical centralized decision algorithm and multiple information filters in parallel on its own states, its teammates' states, and its own states as viewed by its teammates. Under perfect information, the decision architecture allows implicit coordination. Under imperfect information, the estimation of team members' states enables predicting their cost to prosecute new tasks. Thus, the group performance under communication imperfections can be improved. Two different algorithms are proposed for the estimation process. The first is communication efficient, in which asynchronous information updates are sent to the network by individual members based on the value of the information to the rest of the group. The second is computation efficient utilizing synchronous information updates. Taking into account that the plan and plant of each UAV are known to the group improves the overall estimation process. Utilizing the MULTIUAV2 simulation testbed, a Monte Carlo study is presented. The benefit of using the proposed algorithms is shown with regard to the target prosecution rate and the communication bandwidth required for cooperation.
展开▼