Real-Time Optimal Admittance Shaping With Hierarchical Passivity-Based Control for Safe pHRI

Abstract

This article presents an optimal admittance control method for safe and smooth human-robot collaboration, where velocity and jerk are constrained to maintain safety without compromising manipulability. While limiting velocity is essential for safety in physical human-robot interaction (pHRI), it often causes abrupt jerk changes that degrade performance. Resolving this requires a smooth deceleration profile and accurate prediction of the final velocity at the end of deceleration. To address this challenge, an optimal control approach is developed to generate velocity profiles satisfying predefined limits on velocity and jerk. In addition, the overall system's passivity is achieved through an adaptive PD controller that regulates the null-space dynamics of redundant manipulators. This method achieves benefits similar to energy tank approaches while ensuring more reliable passivity without explicit energy management. The proposed method is validated through ISO/TS 9241-411 Fitts' law experiments, demonstrating that this safety-enhancing method maintains operational performance comparable to that of unconstrained controllers. Additional experiments validate the controller's real-time responsiveness, showing its ability to adapt immediately to real-time changes in speed limits and achieve rapid, stable trajectory adjustments. These results highlight the method's applicability to pHRI environments requiring stringent safety and consistent operational performance.