SEANFIS combined with UKF for GPS/MEMS-INS integrated positioning errors prediction

To solve the problem that GPS(global positioning system)/MEMS(micro-electro-mechanical-system)- INS(inertial navigation system) integrated positioning errors accumulate rapidly with time during GPS outages, a kind of SEANFIS (self-evolving adaptive neuro-fuzzy inference system) is proposed to predict the positioning errors of GPS/MEMS-INS integrated navigation system. A hybrid technique combining fading adaptive KF (Kalman filter) and gradient descent algorithm is used to tune the parameters of SEANFIS in real time. Short-term UKF(unscented kalman filter) prediction is combined with long-term SEANFIS prediction effectively, which can predict positioning errors dynamically. The test results show that when compared to traditional ANFIS,SEANFIS has better dynamic adaptability and further improves the positioning accuracy of GPS/MEMS-INS integrated navigation system. Further more, SEANFIS avoids the computation centralization and ensure the real-time performance of the system.