A hybrid type-2 fuzzy logic system and extreme learning machine for low-cost INS/GPS in high-speed vehicular navigation system

Abstract

Due to the combined navigation system consisting of both Inertial Navigation System (INS) and Global Positioning System (GPS) in a complementary mode which assure a reliable, accurate, and continuous navigation system, we use a GPS/INS navigation system in our research. Because of the conditions of navigation system such as low-cost MEMS-based inertial sensors with considerable uncertainty in INS sensors, a highly noisy real data, and a long term outage of GPS signals during our flight tests, we enhance the positioning speed and accuracy by an Extreme Learning Machine (ELM) with the features of excellent generalization performance and fast learning speed. However, the generalization capability of ELM usually destabilizes with uncertainty existing in the dataset. In order to fix this limitation, first, a Type-2 Fuzzy Logic System (T2-FLS) handles the uncertainties in GPS/INS data, and then the final output ends up to the ELM to train and predict INS positioning error. We verify the efficiency of the suggested method in the estimation of speed and accuracy in INS sensors error during GPS satellites outage, particularly in real-time applications with a high-speed vehicle. Then, to evaluate the overall performance of the proposed method, the achieved results are discussed and compared to other methods like Extended Kalman Filter (EKF), wavelet-ELM, and Adaptive Neuro-Fuzzy Inference System (ANFIS). The results present considerable achievement and open the door to the application of T2-FLS and ELM in GPS/INS integration even in severe conditions.