Physics‐Based Torque Prediction Model for Excavating Drums on Granular Soil

ABSTRACT Accurate torque prediction for excavating drums is critical for the design and autonomous control of planetary rovers operating in loose, unstructured regolith. Existing terramechanics models may fail to capture the full dynamics of rotating excavation drums. In this paper, we present a unified, physics‐based torque prediction model with two novel dynamic effects: a longitudinal, velocity‐sensitive bulldozing resistance term that grows logarithmically with forward speed, and a regolith‐accumulation torque that evolves as the drum buckets fill. We validate our approach on a custom four‐wheel holonomic rover excavating Silica Sand No. 5 across multiple test sequences spanning different angular and translational velocities. The model achieves a normalized RMSE generally below 12%, an MAE under 0.15 Nm, and , in nearly all cases, accurately reproducing the temporal evolution toward steady‐state torque plateaus. Its modular structure and reliance on minimal empirical tuning make it suitable for real‐time torque estimation in future autonomous planetary missions.