Design of a Dynamic Trajectory Cooperative Control Model for Autonomous Vehicles Based on Urban Traffic

To address the complexity and real-time challenges of trajectory cooperative control for autonomous vehicles in high-density urban traffic scenarios, a dynamic trajectory cooperative model integrating multi-agent reinforcement learning and distributed model predictive control is proposed<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[1]</sup>. Through multi-source heterogeneous sensor data fusion and spatiotemporal alignment techniques, a dynamic traffic scene representation is constructed<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[2]</sup>; based on an attention mechanism MARL framework, multi-vehicle cooperative intention prediction and conflict resolution are realized, combined with DMPC rolling horizon optimization to generate safe and efficient local trajectories. Simulation experiments show that in a high-density scenario of 40 vehicles/km<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, the trajectory tracking error is less than 0.15m, the conflict resolution success rate exceeds 95%, and communication delay is controlled within 35ms, verifying its robustness and real-time performance in complex dynamic environments<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[3]</sup>, providing theoretical support and technical pathways for urban intelligent transportation systems<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[4]</sup>.