为降低复杂交叉口场景下自动紧急制动（Autonomous Emergency Braking, AEB）系统的碰撞风险并兼顾乘坐舒适性，提出了一种基于运动学能量等效人工势场与模型预测控制（Kinematic Energy Equivalent Artificial Potential Field and Model Predictive Control, KEE-MPC）的分层避撞策略。首先，在上层决策模块引入运动学能量等效人工势场（Kinematic Energy Equivalent Artificial Potential Field, KEE-APF）模型，将路口物理冲突转化为动态碰撞动能评估，并设定双重动态跃度边界机制以实现刚柔控制。其次，设计二维模糊逻辑控制器，以碰撞时间和等效排斥减速度为输入，实时动态调节模型预测控制（Model Predictive Control, MPC）优化框架中状态跟踪与控制增量的权重配比。然后，在底层执行模块构建逆纵向动力学与连续时间比例-积分-微分（Proportional-Integral-Derivative, PID）反馈补偿相融合的双轨结构，精准跟踪期望减速度。最后，依托联合仿真平台开展突发横穿等极限工况测试。结果表明，所提控制策略在全部测试工况中均实现了100%的有效避撞，而传统碰撞时间（Time-to-Collision, TTC）策略避撞率仅为55%；相较于传统控制策略，该策略将最大制动加速度变化率有效约束在6.0~9.1 m/s³，远低于传统策略的30 m/s³；所提策略在满足极限工况避撞安全要求的前提下，改善了车辆的乘坐舒适性，可为复杂交互场景下的自动紧急制动系统多目标优化提供参考依据。

To reduce the collision risk of the Automatic Emergency Braking (AEB) system in complex intersection scenarios and take into account passenger comfort, a hierarchical collision avoidance strategy based on Kinematic Energy Equivalent Artificial Potential Field and Model Predictive Control (KEE-MPC) is proposed. Firstly, the Kinematic Energy Equivalent Artificial Potential Field (KEE-APF) model is introduced in the upper-level decision-making module to transform the physical conflicts at intersections into dynamic collision kinetic energy assessment, and a dual dynamic jump boundary mechanism is set up to achieve rigid and flexible control. Secondly, a two-dimensional fuzzy logic controller is designed, taking collision time and equivalent repulsive deceleration as inputs, to dynamically adjust the weight ratio of state tracking and control increment in the Model Predictive Control (MPC) optimization framework in real time. Then, a dual-track structure integrating inverse longitudinal dynamics and continuous-time Proportional-Integral-Derivative (PID) feedback compensation is constructed in the lower-level execution module to precisely track the expected deceleration. Finally, tests under extreme conditions such as sudden crossing are conducted on a joint simulation platform. The results show that the proposed control strategy achieves 100% effective collision avoidance in all test conditions, while the traditional Time-to-Collision (TTC) strategy only has a 55% collision avoidance rate. Compared with the traditional control strategy, the proposed strategy effectively constrains the maximum acceleration rate of change within 6.0 to 9.1 m/s³, which is much lower than the 30 m/s³ of the traditional strategy. The proposed strategy improves the vehicle's ride comfort under the premise of meeting the safety requirements of collision avoidance in extreme conditions, and can provide a reference basis for the multi-objective optimization of the AEB system in complex interaction scenarios.