Cooperative localization with information-seeking control

We propose a Bayesian method for cooperative localization and control in mobile agent networks. Distributed, cooperative self-localization of each agent is supported by an information-seeking control of the movement of the agents. For cooperative localization, the SPAWN message passing scheme is used. Cooperative control is achieved by maximizing the negative joint posterior entropy of the agent states via a gradient ascent. The localization part of our method provides the control part with sample-based probabilistic information. Simulation results demonstrate intelligent behavior of the agents and excellent localization accuracy.