Centroidal Angular Momentum-Aware Planning for Legged Locomotion Integrating 3D-DCM and Swing-Leg Trajectory Optimization

In humanoid motion planning, centroidal angular momentum (CAM) is often neglected or simplified due to the complexity arising from its nonlinear and non-holonomic nature. As a result, unmodeled CAM effects can compromise center of pressure (CoP) tracking and, consequently, reduce robustness during motion execution. To address this, we propose an online iterative algorithm that computes the induced CAM and updates the corresponding center of mass (CoM) trajectory accordingly. Since leg motion is the primary source of angular momentum in legged locomotion, the proposed framework also optimizes the swing-leg trajectories to generate the desired CAM while respecting kinematic hardware limits. This results in CAM-consistent whole-body motions that enable stable yet faster locomotion with reduced demand for maximum joint velocities. We demonstrate the effectiveness of our approach through experiments and simulations with the humanoid robot TORO in several locomotion scenarios, including balancing, walking, and running.