Abstract

This paper addresses the problem of optimal estimation in a feedback loop where both communication channels, feedback path and control path, are subject to data loss. To deal with measurement loss, the controller embeds a data dropout compensator that replace the missing data with optimal estimates of the measurement. For the case of control signal loss, we consider that the controller sends not only the current control input but also a sequence of future control inputs. Such sequence is stored in a buffer at the actuator side, to employ in case of data loss. To obtain our results we consider that almost all the controller processing is an affine function of the collected data. In this scenario, we obtain optimal estimates of both the plant state and also the buffer state that, given the linear nature of the controller, are computationally inexpensive and easy to implement.