Preload-Dependent Protective Reaction Latency in a Pneumatic Artificial Muscle-Actuated Humanoid Joint

Humanoid robots operating near humans require short protective reaction times in physical human–robot interaction (pHRI). Safety standards distinguish between quasi-static and transient contact. This paper quantifies the reaction timing of a compliant pneumatic artificial muscle (PAM) mechanism under controlled preload conditions. Measurements were performed at 10 N, 50 N, and 100 N preload using synchronised load-cell force, PAM pressure, actuator position, and force-sensitive resistor (FSR) signals. Reaction timing was evaluated relative to the FSR-defined contact onset, at which the controller issued the pressure-release command. The force trace reached its first post-contact peak within 15–20 ms after onset, while the pressure peak occurred within 5–15 ms. A 90% recovery of the post-contact force excursion was achieved within 40–50 ms, whereas the corresponding pressure excursion required 155–180 ms. These timing results quantify reaction latency in PAM-actuated humanoid joints and support multi-modal sensing for robust onset localisation and mitigation monitoring in both ISO/TS 15066 contact types.