Optimal estimation over unreliable communication links with application to cognitive radio

This paper considers state estimation of linear stochastic discrete-time systems through an unreliable network. Packet losses from the sensor to the estimator are assumed to follow a Bernoulli distribution, while information loss is not assumed to be available at the receiver at any time. The optimal estimator provides a satisfactory estimate under unreliable communications as the estimator is used successfully in a linear quadratic Gaussian (LQG) controller to stabilize an inverted pendulum system, and comparison with an optimal estimator for which information loss at the receiver is available. The optimal estimator for a cognitive radio system is derived. Cognitive radio is an emerging technology with many promising applications. Cognitive radio can be considered as a combination of packet loss with and without acknowledgement in the dynamical model. An illustrative example shows that the optimal estimator improves the performance largely.