Prediction of MEMS-based INS Error Using Interval Type-2 Fuzzy Logic System in INS/GPS Integration

In recent years, for reliable, accurate, and robust navigation, Global Positioning System (GPS) and Inertial Navigation System (INS) has been integrated to use their complementary advantages and overcome their drawbacks. Kalman Filtering methods such as Extended Kalman Filter (EKF) have been used for INS/GPS integration widely. The EKF-based navigation systems are complex, and they might not have effective real-time performance, especially with the MicroElectro Mechanical System (MEMS)-based INS when GPS is blocked. To overcome these problems, Artificial Intelligence (AI) based integration was proposed over the Kalman filtering models. Due to the stochastic noise, bias, and drift of the low-cost MEMS-based inertial sensor outputs over time, in this study, we propose an Interval Type-2 Fuzzy Logic System (IT2FLS) to predict the MEMS-based sensor errors in GPS blockage. The IT2FLS can model uncertainty and stochastic noise of both input and training data in complex, noisy environments such as our application. Therefore, we use the IT2FLS to forecast the cumulative INS error during GPS outages to improve the accuracy of the navigation system. The experimental tests show that the IT2FLS has acceptable realtime performance and accuracy in predicting the INS error during the long-term GPS outages.

I. Introduction

Both Global Positioning System (GPS) and Inertial Navigation System (INS) have characteristics that complement each other. Consequently, they have been widely integrated for reliable, precise, and robust navigation systems [1]. Micro-Electro Mechanical System (MEMS)-based INS are broadly used for low-cost navigation systems [2]. The Kalman filtering techniques have been commonly applied to integrate the INS and GPS [3]–[5]. However, the Kalman filtering methods have several issues, such as inefficient realtime performance because of computational complexity, the necessity of finding the representation of each inertial sensor error model, and the weak efficacy and accuracy during the long-term GPS blockage owing to the divergence of the filter from true estimate in high amount of error, specifically when they are applied to MEMS-base INS/GPS integration [2], [6].