Integrated decision-making and path planning framework for autonomous driving in multi-lane obstacle avoidance

Obstacle avoidance in multi-lane traffic scenarios remains a critical challenge for autonomous vehicles, requiring robust decision-making and precise path planning to ensure safety and efficiency in dynamic environments. This paper proposes an integrated framework combining a Time-to-Collision (TTC)-based module for rapid risk assessment and a Large Language Model (LLM)-assisted decision-making module to handle complex situations involving conflicting risks. A novel Velocity-Direction Decomposition (VDD) kinematic model is introduced to address the limitations of classical Longitudinal-Lateral Decomposition (LLD) methods, ensuring smooth and dynamically feasible motion. Model Predictive Control (MPC) is employed to generate collision-free trajectories that respect vehicle dynamics while maintaining stability and passenger comfort. Simulations validate the framework across various scenarios, demonstrating its capability to adapt to diverse traffic conditions, enhance path feasibility, and improve overall system safety and efficiency.