Cooperative receding horizon control for multi-agent rendezvous problems in uncertain environments

We consider the problem of controlling vehicles to visit multiple targets and obtain rewards associated with each target with the added requirement that two or more vehicles are required to be present in the vicinity of each target in order to collect the associated reward. The environment is uncertain and targets may appear or disappear in real time. In view of the problem complexity, we focus on an optimization-based algorithm that makes decisions driven by available information. We develop a cooperative receding horizon controller to maximize the total reward obtained in a finite mission time horizon. We control the motion so as to maximize the expected rewards over a planning window, and show that the resultant trajectory is stationary in the sense that vehicles converge to targets without explicit target assignment.