Terrain Perception and Stiffness Adaptation-Based Attitude Control for Quadruped Robots

This work reports a terrain perception and stiffness adaptive control scheme with no vision. This strategy can enable quadruped robots to operate in a stable posture in unstructured geographic environments. We design a terrain perception controller using Kalman filtering observation technology to accurately estimate the robot's speed, position, and attitude variables. The compensation effect of estimations can provide real-time prediction of the external conditions, suppressing the body's high-frequency jitter in fluctuating terrain and improving anti-interference ability. Yet, a stiffness adaptive controller using impedance theory decreases the actuator's torque deviation and improves the robot's dynamic stability by solving the optimal stiffness change law. The Lyapunov approach proves the stability of the system. In comparative simulations, the proposed method is evaluated against the MIT strategy. Results show that while the MIT method exhibits rapid convergence and effectively reduces initial Euler angle fluctuations, it also introduces increased oscillations during the transient phase. Simulation results and hardware experiments illustrate that the proposed method effectively enhances the robot's terrain adaptability and motion smoothness.