IMU-Error Estimation and Cancellation Using ANFIS for Improved UAV Navigation
Ahmed E. Mahdi, Ahmed Azouz, Ahmed Abdalla, Ashraf Abosekeen
Abstract
The inertial navigation system (INS) is the primary exporter of navigation information in autonomous vehicle (AV) applications. The main component of the INS is the inertial measuring unit (IMU) which is responsible for providing the INS with both angular rates and accelerations. However, the IMU suffers from several types of errors. Some of these errors are deterministic such as bias offset and scale factor while other errors are stochastic such as bias instability, bias drift, and noise. Therefore, the IMUs especially the low-cost micro-electro-mechanical systems (MEMS) provide the INS with drifted raw measurements. Thus, a accumulative error over time affects the INS navigation solution. Recently, the errors related with IMU are modeled and mitigated bu using machine learning (ML) techniques. In this paper, an adaptive neuro-fuzzy inference system (ANFIS) is introduced to improve the effectiveness of low-grade IMUs by estimating and removing the associated errors. A high-fidelity simulated UAV trajectory was utilized to assess the effectiveness of the suggested algorithm. The improvement in the INS navigation solution ia very clear after using the proposed ML-based-ANFIS algorithm with regards to the traditional INS. The solutionis improved by 84.2% and 87.5% in the 2D-positioning and 3D-positioning respectively. Furthermore, 86% and 88.8% in the 2D-velocity and 3D-velocity respectively is achieved when evaluated with regards to the conventional INS solution.