Control of observation systems with application to radar

We formulate the problem of managing sensors as one of optimal control of the resulting state estimator covariance matrix. Originally posed in the early days of optimal control, this problem has received little attention, despite the proliferation of active sensing systems and sensor networks and the resulting need to target observations to obtain useful state estimates while minimizing energy expenditure. We consider the problem where the parameters of an observation system, to be selected from an given set, are controlled in-the-loop. We give a necessary condition for optimality and a numerical scheme for finding these solutions. Finally, using a convenient parametrization of the ambiguity function for a certain class of pulsed radar systems, we apply this theory to optimal tracking of a target with known dynamics.