Advanced Cognitive System for Reliable Lane Switch Prediction in Autonomous Vehicles

The rapid advancements in Autonomous Vehicles (AVs) have significantly heightened the focus on enhancing road safety. While human drivers can predict the behaviour of surrounding vehicles, AVs are yet to fully replicate this cognitive capability, leading to a substantial risk of collisions, particularly during lane-changing maneuvers. This paper introduces the Cognitive Lane Switching Model (CLSM), an advanced system for lane switch prediction. The CLSM comprises the Sequential Modeling Component (SMC), which captures temporal dependencies, and the Probabilistic State Component (PSC), which models latent states, handles state transitions and adapts to emergent traffic patterns. Utilizing Vehicle-to-Everything (V2X) communication metrics, the system dynamically calculates the Time to Lane Switch (TTLS) and the Probability of Lane Switch (POLS). The proposed CLSM demonstrates a lane-switching prediction accuracy of 97.84%, with an average TTLS of 3.72 seconds and a POLS of 0.56. These results highlight the model’s effectiveness in accurately predicting lane switch events.