In order to improve the real-time control ability of biped robot, a real-time obstacle avoidance position control method for biped robot based on reinforcement learning is proposed. Taking the stability of biped walking as the control objective function, the real-time path dynamics model of biped robot is constructed. The acceleration and inertia moment of the robot′s centroid motion are taken as the controlled object. The effective collision sub-model is used to plan the real-time obstacle avoidance path of biped robot, and the collision sub-model and swing sub-model are combined to adjust the error correction parameters of biped robot. The fuzzy reinforcement learning tracking method is used to control the error gain of biped robot, and the real time obstacle avoidance position control of biped robot is realized. The simulation results show that the proposed method can avoid obstacles in real time and improve the adaptive control ability of biped robot.