Research on Stability Optimization of Quadruped Robots in Complex Terrain Under the Improved ZMP Theory

To address the static limitations and lack of real-time terrain feedback in traditional Zero Moment Point (ZMP) theory for quadruped robots operating on dynamically complex terrains, this paper proposes an improved ZMP-based dynamic centroid adjustment method. By introducing dynamic weighting factors that utilize exponential decay functions to adjust joint torque distribution, we balance motion speed and stability requirements. Simultaneously, a real-time terrain inclination estimation model is constructed based on foot-end force sensor data to correct gravitational component deviations in ZMP calculations and reconstruct dynamic support polygons. A hierarchical control architecture is further designed: during dynamic adjustment, Model Predictive Control (MPC) first performs rolling optimization of future centroid trajectories to ensure ZMP remains within the support region, followed by real-time inverse dynamics computation to achieve precise joint torque execution. To validate effectiveness, slope walking and rugged terrain scenarios are tested on the MATLAB simulation platform. Experimental results demonstrate that the improved method reduces ZMP tracking errors compared to traditional ZMP approaches.