A Robust Following Controller for Lower Limb Exoskeletons With a Decoding Human Motion Intention Network Under Small Sample Conditions

摘要

It is still an unresolved challenge to achieve natural and stable human– robot interaction control for lower limb exoskeletons (LLEs). The purpose of this article is to provide a new framework for human–robot interaction control of LLEs. First, we designed the generative adversarial network-based graph transformer (GANGT) for application under small sample surface electromyography (sEMG) dataset conditions. The adversarial learning is utilized to generate featured sEMG data, and graph structure-based temporal and spatial self-attention mechanisms are constructed to capture the local and global features of the sEMG signal during a gait cycle. Second, we designed a robust following controller to control the tracking error of the predicted joint angles and proved its stability through the Lyapunov function. The controller is able to maintain the tracking error within a small bound in the case of unmodeled errors and external disturbances in the dynamical system of the exoskeleton. To validate the effectiveness of the control framework, we conducted a series of experiments with recruited subjects. GANGT demonstrated a better ability to generate hip (RMSE 2.7546<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$^{\circ }$</tex-math></inline-formula>) and knee (RMSE 4.5206<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$^{\circ }$</tex-math></inline-formula>) trajectories in ablation and comparison experiments. The means of the absolute tracking errors for the hip and knee of the exoskeleton under both the level autonomous walking condition and the uphill autonomous walking condition were 0.0846<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$^{\circ }$</tex-math></inline-formula> and 0.1210<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$^{\circ }$</tex-math></inline-formula>, respectively.