A state estimation strategy for a nonlinear switched system with unknown switching signals

"A strategy is presented to estimate the state of a nonlinear autonomous switched system, with no knowledge of the switching signal, except its dwell time. To do so, algorithms to estimate the switching times and the current mode of the system are developed. The estimation of the switching times is based on approximating the second ( generalised) derivative of the output of the system via a convolution of this signal with a suitable function and on detecting the corresponding spikes. To estimate the modes, a scheme based on the use of a bank of observers (one for each mode) and of a bank of subsystems (for each step of the estimation process a suitable subset of the subsystems of the switched system) is developed. The algorithms run regardless of the state observer model, as long as its output error norm decays exponentially with a controlled decay rate."