Disturbance-Aware Preview Control for Humanoid Robot Towing of Heavy Wheeled Loads with Traction Force Compensation

Most existing humanoid robots possess a payload capacity below 10 kg, which is insufficient for practical material transport tasks. A natural engineering solution is to enable humanoid robots to tow wheeled trailers or hand carts, transferring the vertical load to the wheeled carrier while the robot provides only horizontal traction. This approach is mechanically feasible, low-cost, and compatible with existing logistics equipment. However, the control challenges have not been adequately addressed in the literature. The traction force exerted by the towed load acts as a persistent disturbance on the robot’s center of mass (COM), degrading zero-moment point (ZMP) tracking and compromising gait stability. This paper proposes a disturbance-aware preview control framework for humanoid robots towing heavy wheeled loads. First, the measured traction force is incorporated as a feedforward term into the linear inverted pendulum model (LIPM) to compensate for its effect on gait symmetry. Second, a disturbance-aware preview controller is designed that optimally generates COM trajectories under sustained traction forces by incorporating both ZMP reference preview and traction force preview into a unified optimal preview control framework. The method integrates COM dynamics, traction-force closed-loop compensation, and ZMP reference preview into a single analytical optimal preview control framework, with gait-phase scheduling incorporated through the footstep-based reference trajectory. Simulation results demonstrate that the proposed method reduces ZMP tracking error by 13.5% and COM jerk by 37.7% compared to standard preview control, while maintaining robust walking stability under traction forces exceeding about 180 N.