Efficient and Robust Standing Postures of Quadruped Robots

Inspired by the natural posture adjustments of animals under external loading, this article presents an optimization‐based framework for minimizing joint torques of quadruped robots in a standing posture subjected to external payloads. By exploiting kinematic and static relationships, a torque‐driven control strategy is formulated to compute minimized torque solutions for the redundant quadruped system. Beyond direct torque optimization, the framework adapts the posture to a nearby optimal pose, further reducing joint torques. This approach achieves up to 50% torque reduction compared to position‐based control, particularly under heavy loading conditions. To enhance robustness in real‐world environments, the study proposes a robust posture optimization considering a set of external loads from all possible directions, thereby preventing postures that are resilient in one direction but vulnerable to disturbances in others. The resulting robust posture search identifies configurations that minimize joint torques under worst‐case loading scenarios. The proposed methodology is validated through simulations in Gazebo and hardware experiments on the Unitree Go1 quadruped. Results demonstrate accurate estimation of external loads using the static model and consistent, significant torque reduction across diverse conditions, confirming the effectiveness and practicality of the approach.