Application of federated particle filter fusion and fault tolerance in integrated navigation system
Ziyu Li, Duan Chenglin, Chang Bing, Yuan Xudong · 2017
In multi-sensor integrated navigation systems with non-linear subsystems and non-Gaussian noise, errors, even failures, exist in global state estimation obtained by federated Kalman filter. χ2residual test and χ2state test are invalid in nonGaussian models. Since likelihood function can be estimated by particle filter (PF), by combining PF with federated Kalman filter, this paper proposes a likelihood detection method based on federated PF technology. The method is applied in JIDS/SINS/GPS integrated navigation system (INS). Simulation in non-Gaussian noise environment shows that the method is effective in fault detection, isolation and recovery. It improves positioning precision and reliability in INS.