Abstract: An adaptive collision avoidance control strategy based on working condition identification is proposed to solve the problem of active collision avoidance under high-speed emergency conditions. According to the real-time traffic environment information and vehicle states information, a classification method of emergency conditions avoidance mode is constructed. The method divides the emergency mode collision avoidance mode into three modes, braking collision avoidance, steering collision avoidance, and coordinated collision avoidance. For braking collision avoidance mode, a longitudinal braking collision avoidance strategy considering road friction conditions and the comfort of the driver and passengers is designed to solve the problem of braking collision avoidance. A collision avoidance strategy based on polynomial path planning is used to solve the collision avoidance problem in the steering collision avoidance mode. For braking and steering coordination collision avoidance mode, a data-driven self-learning coordinated control strategy is designed to address the coordinate collision avoidance problem. The expected output of different control strategies is tracked by the PID lower controller to complete collision avoidance. The Simulink-Carsim vehicle emergency collision avoidance control joint simulation platform is built in Matlab /Simulink environment. Based on the platform, a virtual test of various working conditions is carried out to verify the real-time and effectiveness of the control system. The results show that the control system can automatically and effectively identify what kind of collision avoidance operation should be adopted in the current emergency conditions, and the stability of the vehicle can be ensured while completing the collision avoidance.