Data Validity Analysis Based on Reinforcement Learning for Mixed Types of Anomalies Coexistence in Intelligent Connected Vehicle (ICV)

Abstract

Compared with traditional anomaly analysis, intelligent connected vehicle (ICV) data validity analysis is faced with a variety of data anomalies, including sensor anomalies, driving behavior anomalies, malicious tampering, and so on, which eventually leads to anomalies in the data. How to integrate the vehicle moving characteristics, driving style, and traffic flow conditions to provide an effective data detection method has become a new problem in the field of intelligent networked vehicles. Based on ICV data, a particle swarm optimization data validity detection algorithm (TE-PSO-SVM) was proposed by combining driving style and traffic flow theory to realize the effective detection of driving data. In addition, aiming at the problem of mixed types of anomalies in complex scenes, a model pool is constructed, and a model selection algorithm based on reinforcement learning (RLBMS) is proposed. Experiments on the real data set HighD show that RLBMS has a better detection effect in complex scenes of mixed types of anomalies.

Single and Mixed Sensory Anomaly Detection in Connected and Automated Vehicle Sensor Networks

“…Connected and automated vehicle (CAV) facilities create a safer, more efficient, and more sustainable transportation system for consumers [1][2][3]. The development of CAVs is closely linked to the development of other emerging technologies, such as artificial intelligence (AI) and the Internet of Things (IoT), which are expected to transform various industries and consumer experiences [4,5]. CAVs can be categorized into different levels of automation, ranging from partially automated vehicles with features such as adaptive cruise control and lane-keeping assistance to fully autonomous vehicles that can operate without human input [6][7][8].…”