Energy efficient state estimation through stochastic optimization

We address the design of energy efficient state estimation in wireless sensor networks satisfying a desired average accuracy constraint over a time-varying channel. We propose a new radio resource allocation policy based on Lyapunov drift stochastic optimization to be used with a standard Kalman filter estimator. The salient feature of the framework is that it can achieve arbitrarily close to optimal power efficiency over time without requiring knowledge of the channel statistics or future events. Asymptotic optimal performance is achieved at the expense of an increase in latency for the system to converge to the desired estimation accuracy. The explicit trade-off is governed by a tunable parameter V. This work unifies notions of estimation and network control optimization.