In view of the difficulty in designing multi-objective planning algorithms during emergency steering for collision avoidance and the complexity and variability of the number and location of obstacles, a hierarchical decision-making planning algorithm combining sampling and optimization is proposed. Considering environmental and kinematic constraints under structured roads, a variant of the A* algorithm is used to establish the surrounding environmental potential field and compute the kinematic cost using a fifth-degree polynomial. Travel lanes are established based on the coarse trajectories, and the quadratic planning problem is solved using the segmentation-plus-acceleration method to obtain smooth paths so as to ensure comfortable path planning, while guiding the vehicle back to the center of the road. The results of simulation tests and real vehicle tests show that the proposed scheme can flexibly complete the decision planning tasks according to different obstacles and achieve emergency collision avoidance.