Risk-sensitive sensor scheduling for discrete-time nonlinear systems

We consider a discrete time, partially observed risk sensitive optimal control problem in which observations of the state are available from a finite number of different measurement sources. At each time step, one sensor can be chosen. We study the problem of finding the sensor schedule and the system input so that a risk-sensitive cost is minimized. For the general nonlinear case, the problem is generally infinite dimensional. In the case of linear systems however, the information state is finite dimensional, and a finite dimensional suboptimal strategy is presented when the cost is quadratic