Potential sources of sensor data anomalies for autonomous vehicles: An overview from road vehicle safety perspective

Xiangmo Zhao, Yukun Fang, Haigen Min, Xia Wu, Wuqi Wang, Rui Teixeira

Abstract: Outstanding steps towards intelligent transportation systems with autonomous vehicles have been taken in the past few years. Nevertheless, the safety issue in autonomous vehicles is critical and remains to be fully solved. Sensor data provide information about the internal status of the system and the impact of its external environment, where the occurrence of sensor data anomalies indicates the existence of potential safety risks. Therefore, in this work, a taxonomy for potential sensor data anomaly sources from the perspective of road vehicle safety is proposed, motivated by the lack of a unified comprehensive taxonomy of sensor data anomaly identification for autonomous vehicles. In this context, sources are divided into; 1) fault or failure of the components or subsystems; 2) failure of the adaptability to the external environment; 3) cyber-attacks; and 4) faults or design deficiencies of sensors. Based on the taxonomy proposed, related works, and in particular, countermeasures for the four potential sources of sensor data anomalies in autonomous vehicles are then reviewed. In the context of providing a comprehensive discussion, other taxonomies of potential sources causing sensor data anomalies for autonomous vehicles and the issue of interpretability of sensor data anomalies are also discussed, providing insight into the strengths of the proposed taxonomy.