AI-Empowered Trajectory Anomaly Detection and Classification in 6G-V2X

Abstract

The immense growth of Autonomous Vehicles (AVs) and networking technologies have paved the way for advanced Intelligent Transportation Systems (ITS). AVs increase data demands from in-vehicle users, which pose a significant risk to the vehicular trajectory data and are extremely vulnerable to security threats. It is challenging to describe and detect the trajectory anomalies in urban motion behavior due to the enormous coverage and complexity of ITS in the V2X environment. Most existing systems rely on a restricted number of single detection strategies, such as determining frequent patterns and have limited accuracy in detecting anomalous trajectories. However, they focus only on outlier detection, failing to consider different patterns of anomalous trajectories. This paper proposes Efficient Trajectory Anomaly Detection and Classification (ETADC) framework in a 6G-V2X environment. The proposed ETADC framework employs the Deep Deterministic Policy Gradient algorithm (DDPG) to improve accuracy and efficiency by analyzing multiple strategies, namely driving speed, driving distance, driving direction, and driving time. The result analysis shows that the proposed ETADC technique outperforms the existing systems by 97% accuracy.