Physical Human–Robot Interaction of Dual-Arm Manipulators With Human Intention Estimation and Two-Level Admittance Adaptation

Physical human–robot interaction (pHRI) is important for seamless collaboration, improved efficiency, and flexibility whilst robots perform their tasks. Unlike many studies that have focused primarily on single-arm manipulator pHRI, this study specifically addressed the pHRI challenges associated with dual-arm manipulator systems, which can be more difficult problems to solve owing to the inclusion of force couplings, kinematic constraints, uncalibrated trajectory deviations, and uncertainties related to the controlled object. Specifically, this study proposed a human-intention estimation and two-level admittance adaptation control scheme for dual-arm manipulators pHRI, to estimate unknown human motion intent, reduce human effort, and implement effective force control under conditions of uncertainty. An adaptive neural network was employed to estimate the unknown human motion intent, enabling two manipulators to actively follow human movements. An object-level force feedback variable damping admittance controller (FFVDAC) was proposed to enhance pHRI compliance and minimize human effort by adjusting the damping parameters. In addition, a novel manipulator-level neural-network-based variable stiffness admittance controller (NNVSAC) was proposed to effectively control the internal forces under conditions of uncertainty. The stability of the object-level FFVDAC and manipulator-level NNVSAC were demonstrated through the Routh criterion and Lyapunov stability analysis, respectively. To confirm the feasibility and superiority of the proposed pHRI control scheme, a series of pHRI and dual-arm manipulators cooperation experiments were conducted using two redundant manipulators. The results clearly demonstrated that the proposed control scheme can successfully reduce human effort and internal force oscillations. Moreover, comparative analysis revealed the proposed algorithm's superior performance.