Multi-object two-agent coordinated search

In this paper, we discuss stochastic control approaches to sensor control problems for the purposes of locating multiple objects in a compact domain using two agents. Unlike most previous work, we consider measurement models that do not focus measurements on individual object locations, but instead obtain indicators of the presence of objects in a chosen region. By changing the region which they observe, sensing agents collect enough information to locate the objects of interest. Our work extends similar results for the case of one sensing agent and one hidden object to the case of twp sensing agents and multiple objects. We derive an optimal multistage sensing policy for the case of two agents, and provide an explicit construction for determining adaptively the sensing actions for each agent under the assumption that they share the sensed measurements. We discuss extension of the results to multiple agents, and illustrate the performance of our proposed policy in simulation, comparing to alternative policies based on spatial domain partition for each agent.