Mode discernibility and bounded-error state estimation for nonlinear hybrid systems

Abstract : State estimation is a key engineering problem when addressing control or diagnosis issues for complex dynamical systems. The issue is still challenging when the latter systems must be modelled as hybrid discrete-continuous dynamics, which is true for many complex and safety-critical systems. In this paper, we investigate nonlinear hybrid state estimation in a bounded-error framework using reliable and robust methods. We first establish a testable condition for current mode location discernibility. Then we build our hybrid state estimator which relies on a prediction-correction approach. An illustrative example is presented.