Abstract: A robust control algorithm is presented for the research platform of humanoid walking robot in order to realize dynamic equilibrium autonomously with real-time performance in the presence of external disturbances within a certain range.The robot dynamic walking process is depicted as an anthropomorphic planar mechanical system of nine links.The dynamics model of robot walking is constructed based on the Coriolis vector etc.,and the constraint conditions for asymptotic stability are obtained by means of non-linear compensation.Moreover,an ideal Lyapunov function is derived whose parameters are optimized with GA(genetic algorithm),and a real-time robust control algorithm is designed.The angular displacement error of each robot joint is calculated by simulation.The results show that the ZMP(zero moment point) is always within the range of supporting foot,and the projection of gravity center trajectory on the ground is a sine curve.The presented algorithm is applied to walking control of a real humanoid robot.The experiment results show that the robot can get to the dynamic stability autonomously in short executing epochs in the presence of external disturbances within a certain range.