Event-triggered particle filtering and Cramér–Rao lower bound computation

In this article, an event-triggered particle filtering method is presented to estimate the states of stochastic nonlinear systems with the ultimate goal to reduce the information exchange in networked systems. In the event-triggered estimation, measurements are transferred to an estimator only if certain event conditions are satisfied. Using these event-triggered measurements leads to non-Gaussianity of the conditional posterior distribution in minimum mean square error estimators even in the presence of Gaussian process and measurement noises. Therefore, in this article, a particle filter–based method is employed to solve the non-Gaussianity issue in nonlinear systems due to event-triggered measurements. In the proposed scheme, when no information is sent to the estimator, particles weight update role is modified according to the event-triggering probability density function. To evaluate the performance of the proposed state estimation scheme, the conditional posterior Cramér–Rao lower bound is obtained using Monte Carlo simulations. The bound is also computed for nonlinear Gaussian systems with a Gaussian event-triggering mechanism as a special case. Finally, the efficiency of the proposed method is demonstrated for a networked interconnected four-tank system through simulation and a comparison study is also provided.