关键词: 智能车辆；模型预测控制；局部路径规划；紧急避障；自适应控制；神经网络

Abstract: The application of model predictive control (MPC) based local path planning algorithm in the field of autonomous driving is becoming increasingly widespread, but this method is prone to planning failure and poor obstacle avoidance effect caused by external parameter changes. To address this issue, an objective function is proposed that includes trajectory deviation, front wheel deflection angle, and obstacle avoidance function. Simulation analysis is conducted under the constraint conditions, and a nonlinear model is established that includes vehicle speed, adhesion rate, and obstacle avoidance weight to optimize local path planning and obstacle avoidance function. By designing three controllers, namely polynomial fitting, adaptive fuzzy logic, and back propagation (BP) neural network, adaptive control of obstacle avoidance weights has been achieved. Build a Simulink/CarSim joint simulation platform to verify the effectiveness of local path planning and obstacle avoidance functions for three controllers. The results show that the BP neural network controller has the best comprehensive performance and good adaptability to different road adhesion conditions, vehicle speeds, and expected paths.

Key words: intelligent vehicle; model predictive control; local path planning; emergency obstacle avoidance; adaptive control; neural network