Abstract
Friction in the joints of robots equipped with harmonic drives exhibits complex dependencies on temperature, load, and velocity. Accurate friction models are critical for designing high-performance model-based controllers that maintain energy efficiency during repetitive operations, ensure safe physical interaction in human-robot collaboration, and enable reliable performance across varying environmental and operational conditions. The majority of the existing works emphasize velocity-dependent friction only. To account for explicit thermal and load effects, we first provide a review and comparison of existing analytical and data-driven models. Next, narrowing our focus on thermal aspects, we numerically and experimentally show the significance of capturing temperature variations in friction modeling for model-based control in robotic manipulation.