Experimental Study under Real-World Conditions to Develop Fault Detection for Automated Vehicles

Automated vehicles can contribute to the improvement of transportation through their high capacity, increased safety, low emission and high efficiency. However, unstable conditions of automated mobile systems, which include automated vehicles and mobile robots) can cause serious problems, andthus, automated mobile system requiresto be highly reliable. The objective of this research is to develop on analgorith mfor detection faults (unstable condition) in an automated mobile system and to improve the overall reliability of this system. In this study, we in itially stored and updated a few patterns of data constellations under normal and unstable conditions for fault identification through real-world experiments. Multiple experiments were performed in a public urban area (with course distance per set beingapproximately1.1(km)), where several pedestrians, bicycles, and other robots were also present. The method used for detecting faults utilizes Mahalanobis distance, correlat ion coefficient, and linearization in order to enhance the accuracy of detecting faults;further, because real-world experimental conditions vary frequently,it is essential for the proposed method to be robust undervarious conditions. The main feature of this study is that it involves the use of experimental results obtained under real-world conditions, to develop a fault detection algorithm and evaluate its validity. In addition, simu lations were performed using the real-world experimental data, wh ich includes newly logged experimental data after the algorithm was developed in order to evaluate the validity of the proposed algorithm. The simulat ion results show that the proposed algorithm detects faults accurately, thus, they prove its validity.