Optimal Sensor Location Design for Tracking Agents with Uncertain Dynamics

In surveillance and monitoring systems with multiple sensors at spatially distributed locations, the locations of the sensors crucially affect the information-content of the sensor readings and thus the quality of the estimates obtained. This paper deals with the problem of choosing the sensor locations for estimating processes with nonlinear and uncertain dynamics that can be represented as Markov chains. We propose an information-theoretic metric to measure the performance of a given set of sensor locations that is independent of the estimation algorithm used. We derive bounds on the metric that makes the methodology computationally tractable and scalable to large problems. Numerical examples are provided to illustrate the designs obtained.