Dynamic, optimal sensor scheduling and value of information

In this paper, we present a method for the optimal state estimation of a linear system, observed by various dynamically schedulable and distributed sensors. We consider this problem to be a problem of information fusion, where the information is obtained from the sensors in exchange of scheduling and sensor operational costs. Sensors over a distributed network can work together efficiently in order to maximize the overall network performance. We consider several costs related to sensor scheduling such as the continuous cost for keeping the sensor on, and sensor switching cost for turning a sensor on from off state and vice versa. Our goal is to estimate the state as closely as possible while minimizing the scheduling cost. We incorporate the error in estimation as a cost in the optimization function. The resultant problem is then converted into an optimal control problem, where optimal scheduling is obtained by pointwise maximizing the Hamiltonian of the system.