Optimal Sensor Scheduling Under Intermittent Observations Subject to Network Dynamics

Motivated by various distributed control applications, we consider a linear system with Gaussian noise observed multiple sensors which transmit measurements over dynamic lossy network. We characterize the stationary optimal sensor scheduling policy for finite horizon, discounted, and long-term average cost problems show that value iteration algorithm converges to solution of problem. further suboptimal policies provided rolling horizon truncation also guarantee stability provide near-optimal cost. Lastly, qualitative characterizations multidimensional set measurement loss rates is stabilizable static network, thus extending earlier results on intermittent observations.