Persistence Filters for Estimation: Applications to Control in Shared-Sensing Reversible Transducer Systems

Abstract

We investigate state observer and feedback control design for systems with state- and time-dependent control or measurement gains. In this framework, we look at reversible transducers that are continually switched between the actuation and sensing modes at some prespecified schedule. Design and analysis of stable state-observers and feedback controllers for these classes of switched/hybrid systems are significantly complicated by the fact that, at any given instant of time, the overall system loses either controllability (during the sensing phase) or observability (during the actuation phase). In this work, we consider systems with scalar time-varying measurement gains and provide a novel observer construction that guarantees exponential reconstruction of state estimates to their true values. We go a step further to derive an exponentially stabilizing controller design that uses the state estimates resulting from our observer. This amounts to the establishment of a rather remarkable separation property of the control design. These developments hinge on a rather mild technical assumption, which can be interpreted for the reversible transducer problem as a persistent dwell time for both the sensing and actuation modes. An important feature here is that the convergence rate can be specified to any arbitrary value. Our theoretical results are validated through numerical simulations of challenging test-cases that include open-loop unstable systems. The paper also illustrates potential for nonlinear extensions of the observer based control design by considering an interesting special case.