An Integrated Localization System with Fault Detection, Isolation and Recovery for Autonomous Vehicles

Since the effect of faulty measurements on the integrated localization of autonomous vehicle is catastrophic, the combination of Fault Detection, Isolation and Recovery (FDIR) and localization has become a major trend to improve the accuracy and reliability of localization for autonomous vehicles. This paper proposes new test variables on the basis of Extended Kalman Filter (EKF) and Gaussian distribution to detect faults. A new isolation method is designed to locate the faulty source and a two-way interaction mechanism is built to achieve fault recovery when there is a fault. Furthermore, we apply the FDIR in an integrated localization system. Due to the lack of public datasets, we collected multiple sets of data (with and without faults) with our autonomous driving platform 'Pioneer'. Four novel metrics for evaluating FDIR algorithms are defined and corresponding comparisons are carried out. In the ablation experiment, we verify the importance and influence of modules of FDIR for localization. Experimental results demonstrate the effectiveness of the proposed framework. And 'Pioneer' won the championship of the China Intelligent Vehicle Future Challenge in 2019 and 2020.