This project will develop an information-theoretic framework for the planning and control of robotic sensor networks. Concepts from distributed sensor networking, cooperative control, networked communication, sensor fusion, and information theory are combined into a single framework that incorporates both communication and sensing objectives. This new framework enables robust intelligence in robot teams through reasoning about the combined effects of sensor uncertainty and wireless networked communication on model-based active sensing. Communication models move beyond line-of-sight or proximity constraints to consider the dependency of channel capacity on relative separation, asynchronous communication, and interference or noise that cannot be modeled. Sensing strategies will perform distributed decision-making over finite planning horizons and will integrate new methods to achieve desired levels of uncertainty rapidly. Research thrusts include: i.) information-theoretic formulation of robot sensor network planning and control; ii.) distributed optimization that combines predefined maneuvers with random search techniques; iii.) robustness to sensor uncertainty and radio noise using receding horizon control and multivariable extremum seeking; and iv.) experimental validation. Education activities will address the leakage of students from the engineering and technology pipeline by focusing on the engineering challenges, scientific questions, and societal benefits associated with the development of robot sensor networks.
