A resilient Extended Kalman Filter for discrete-time nonlinear stochastic systems with sensor failures

Missing sensor data is a common problem which severely influences the overall performance of today's dataintensive applications. In order to address this important issue, a resilient Extended Kalman Filter is proposed for discrete-time nonlinear stochastic system and measurement equations with sensor failures and random gain perturbations. The failure mechanisms of multiple sensors are assumed to be independent of each other with different failure rates. A generalized Extended Kalman Filter is designed to have robustness against sensor failures and resilience against random perturbations in the filter gain. Lorenz oscillator, a benchmark nonlinear chaotic system, is used to demonstrate the effectiveness and resilience of the proposed approach.