Z.K. Chang, Y.W. Chen, X. Feng

Pages: 65-82

Abstract
To enhance safety and smoothness of autonomous vehicle trajectory planning, a dynamic game algorithm-based method for autonomous vehicle trajectory planning is proposed. This approach constructs a comprehensive model encompassing joint vehicles in a lane-changing scenario, integrating vehicle kinematics and dynamics models to accurately depict the interaction behavior between the cutting-in vehicle and the target vehicle. Leveraging game theory, a non-cooperative dynamic game model is formulated to derive the optimization function tailored to the game benefits' decision-making objective. The particle swarm optimization algorithm is employed to solve the objective function and ascertain the optimal lane-changing decision. Subsequently, the state positions corresponding to various action combinations under this strategy are determined, facilitating trajectory planning for unmanned autonomous vehicles. The results demonstrate that the proposed method not only guarantees a minimum safe distance but also ensures a relatively low curvature change rate for the planned trajectory, staying below 0.1. This method holds promising application prospects and significant value in advancing the field of autonomous driving by enhancing trajectory planning accuracy and safety.
Keywords: dynamic game theory; autonomous driving of unmanned vehicles; trajectory planning; particle swarm optimization algorithm; reinforcement learning algorithm