Feedback control over lossy channels: Optimal estimation considering data-loss compensation strategies

Abstract

In this paper we investigate the optimal estimation in feedback loops where communication links are lossy channels. To deal with data loss in the feedback path, the controller embeds a data dropout compensator that replaces the missing measurements with optimal estimates. For control signal loss, we consider a strategy where the controller sends not only the current control input but also a sequence of future control inputs, which is stored in a buffer at the actuator side, to employ in case of data loss. We restrict our attention to controllers in which almost all processing is an affine function of the available data. In this scenario, we obtain optimal estimates of the plant state but also of the buffer state. Given the linear nature of the controller, the optimal estimator is simple to implement. Also, we provide the stationary optimal estimator which is an even simpler and computationally inexpensive alternative.