Fault Tolerant Multi-Sensor Federated Filter for AUV Integrated Navigation

Abstract

Localization and autonomous navigation with high precision, high reliability and low cost is an important part of Autonomous Underwater Vehicle (AUV). Although using more accurate inertial navigation system can improve the accuracy, it will make the cost of navigation system very expensive, and a single sensor cannot cope with the complex environment that AUV is facing. By utilizing the complementary and redundant characteristics of various sensors (SINS/GPS/DVL/AHRS), the improved federated filter can achieve higher navigation performance and stability than using a single system alone. In practical applications, low-cost GPS usually has short-term uncertain abnormal jumps, which will affect the filtering performance of federated filter to a great extent. The traditional federated algorithm of single reference system does not have fault tolerance. At the same time, there exists the problem of poor fault tolerance of filters with feedback and poor precision of filters without feedback. In view of the above problems, this paper proposes a multi-sensor redundant integrated navigation system based on data preprocessing, which improves the stability and accuracy of the federated system and is more suitable for engineering applications.