Event-Triggered State Estimation: Experimental Performance Assessment and Comparative Study

In this brief, an experimental and comparative study of event-based state estimation is performed for some typical event-triggering conditions. The experiments are performed on a magnetic-brake loaded permanent-magnet dc torque motor system. The comparisons are performed from three aspects: 1) performance under different average sensor-to-estimator communication rates; 2) effect of inaccurate estimates of the noise covariance matrices; and 3) computation complexity. The comparative results show that the innovation-based schedules are relatively superior to the 'send-on-delta' schedules in terms of estimation quality, and the deterministic schedules have enhanced estimation performance compared with stochastic counterparts. The estimators considered are sensitive to the estimates of the noise covariance matrices, and retain a similar level of computation complexity as that of the standard Kalman filter.