Wavelet-adaptive neural subtractive clustering fuzzy inference system to enhance low-cost and high-speed INS/GPS navigation system

Abstract

The combined navigation system consisting of Global Positioning System (GPS) and Inertial Navigation System in a complementary mode assures an accurate, reliable, and continuous positioning capability in the navigation system. Because of problems such as dealing with a low-cost MEMS-based inertial sensors having a high level of uncertainty and imprecision, stochastic noise, a high-speed vehicle, high noisy real data, and long-term GPS signal outage during the real-time flight test, the advantage is taken for some approaches in different steps: (1) utilizing discrete wavelet transform technique to enhance the signal-to-noise ratio in raw and noisy inertial sensor signals and attenuate high-frequency noise as a preprocessing phase to prepare more accurate data for the proposed model and (2) employing adaptive neural subtractive clustering fuzzy inference system (ANSCFIS) which combines and extracts the best feature of adaptive neuro-fuzzy inference system (ANFIS), and the subtractive clustering algorithm with fewer rules than the ANFIS method, aiming to improve a more efficient, accurate, and especially a faster method which enhances the prediction accuracy and speeds up the positioning system. The achieved accuracies for the proposed model are discussed and compared with the extended Kalman filter (EKF), ANFIS, and ANSCFIS which are implemented and tested experimentally using a high-speed vehicle in three GPS blockages. The proposed model shows considerable improvements in high-speed navigation using low-cost MEMS-based inertial sensors in case of long-term GPS blockage.