Learning hidden Markov models for linear Gaussian systems with applications to event-based state estimation

Abstract This work attempts to approximate a linear Gaussian system with finite-state hidden Markov model (HMM), which is found useful in dealing challenges designing networked control systems An indirect approach developed, where state-space (SSM) firstly identified for and the SSM then used as an emulator learning HMM. In proposed method, training data HMM are obtained from generated by through building quantization mapping. Parameter algorithms designed learn parameters of HMM, exploiting periodical structural characteristics The convergence asymptotic properties analyzed. learned using applied event-triggered state estimation, numerical results on estimation demonstrate validity algorithms.