Physical Human-Robot Interaction Control of an Upper Limb Exoskeleton with a Decentralized Neuro-Adaptive Control Scheme

TL;DR

This paper presents a decentralized neuro-adaptive control scheme for an upper limb exoskeleton to ensure safe and stable physical human-robot interaction, accounting for uncertainties and human forces, validated through experiments.

Contribution

The paper introduces a novel decentralized neuro-adaptive control strategy with HEF estimation for safe, stable pHRI in high DoF exoskeletons, validated by experimental results.

Findings

Proposed controller outperforms PD and PID controllers in stability and safety.

Experimental validation across different velocities and users demonstrates robustness.

Controller effectively estimates human forces and adapts to uncertainties.

Abstract

Within the concept of physical human-robot interaction (pHRI), the most important criterion is the safety of the human operator interacting with a high degree of freedom (DoF) robot. Therefore, a robust control scheme is in high demand to establish safe pHRI and stabilize nonlinear, high DoF systems. In this paper, an adaptive decentralized control strategy is designed to accomplish the abovementioned objectives. To do so, a human upper limb model and an exoskeleton model are decentralized and augmented at the subsystem level to enable a decentralized control action design. Moreover, human exogenous force (HEF) that can resist exoskeleton motion is estimated using radial basis function neural networks (RBFNNs). Estimating both human upper limb and robot rigid body parameters, along with HEF estimation, makes the controller adaptable to different operators, ensuring their physical safety. The barrier Lyapunov function (BLF) is employed to guarantee that the robot can operate in a safe workspace while ensuring stability by adjusting the control law. Unknown actuator uncertainty and constraints are also considered in this study to ensure a smooth and safe pHRI. Then, the asymptotic stability of the whole system is established by means of the virtual stability concept and virtual power flows (VPFs) under the proposed robust controller. The experimental results are presented and compared to proportional-derivative (PD) and proportional-integral-derivative (PID) controllers. To show the robustness of the designed controller and its good performance, experiments are performed at different velocities, with different human users, and in the presence of unknown disturbances. The proposed controller showed perfect performance in controlling the robot, whereas PD and PID controllers could not even ensure stable motion in the wrist joints of the robot.