A Self-Fault Diagnosis Framework for Sensors of Connected and Automated Vehicles with Dynamic Environmental Impact Quantification

Yukun Fang, Hai-Gen Min, Xiaoping Lei, Xiangmo Zhao

Abstract: Connected and automated vehicles (CAVs) are revolutionizing the development of transportation, while reliability and safety issues for CAVs remain to be improved. Sensor data are the observations of the internal state of the CAV and its external environment, and fault diagnosis for the sensor system aims to provide available information about the operation status of the sensors to the decision-making unit, avoiding potential risks. This paper presents a self-fault diagnosis framework for CAV sensors with environmental impact quantification. We attribute the occurrence of abnormal sensor data patterns to four potential sources, i.e., the sensor itself, the external environment, the cyber-attack, and the measured object, and this paper emphasizes the former two. Specifically, a residual consistency checking algorithm based on the sensor redundancy is proposed to quickly detect and isolate the failed sensor, and the spectral entropy of each sensor data is calculated to assess their health status. Besides, measurement uncertainty is dynamically calculated to evaluate the measurement quality of the sensor system, and the environmental impact on the measuring process is then quantified. Real field experiments were implemented, and experiments using the data collected from the real CAV platform validated the effectiveness of the proposed methods.