Abstract: A vision-free terrain perception and stiffness adaptive control scheme is proposed to address the difficulty of stable motion of quadruped robots in unstructured terrain. A robot state observer is designed using a Kalman filter, to obtain accurate estimates of the robot states. Then, a terrain perception based controller is designed using the estimated robot states to suppress high-frequency jitter of the body in fluctuating environments and improve anti-interference ability. In addition, a stiffness adaptive controller is designed based on impedance theory, to improve the convergence speed of attitude deviation and enhance the dynamic stability of the robot by solving the optimal stiffness change law. The Lyapunov theory is applied to analyzing the system stability. The simulation results and prototype experiments show that the proposed method significantly enhances the terrain adaptability and motion smoothness of the robot.