Humanoid Robot Jumping Control Based on Centroid Trajectory and Sparse Key Limb Pose Guidance

Humanoid robot jumping control is a challenging problem due to its strong nonlinearity and randomness. Although precise Trajectory Optimization (TO) can provide high-quality reference trajectories, its computational cost is prohibitively high, and it is difficult to adapt flexibly to the robot's actual state. In this paper, we propose a jump control framework guided by the centroid and sparse key limb pose trajectories, aimed at enhancing the real-time performance of TO for humanoid robot and achieving high-performance whole-body coordination control. Firstly, the complex dynamics model of the humanoid robot is simplified, and a Single Rigid Body Model (SRBM) is constructed. Using the SRBM, the time series and centroid trajectory of the jumping motion are optimized by incorporating the target foot landing position and complete foot contact force constraints. Next, the sparse whole-body motion trajectory is solved based on the centroid trajectory, extracting the pose of key limbs as the guiding trajectory for the humanoid robot during the jumping motion control process. Finally, based on the dynamic characteristics of each phase in the jumping motion, a Constraint-based Hierarchical Inverse Dynamics Whole Body Controller (CHID-WBC) is constructed to control the jumping motion of the humanoid robot. Simulation and experimental results demonstrate that the proposed control framework effectively enables whole-body coordinated jumping motion of the humanoid robot G1 across multiple operating conditions, validating its effectiveness.