Design and assessment of a control framework for partial assistance on a robotic platform with the Exo-H3 exoskeleton for human gait rehabilitation 

Abstract

Background Rehabilitation of the human gait with robotic assistance requires modulating the human-exoskeleton interaction in such a way that the desired level of assistance is controlled. For this purpose, it is necessary to assess the assistance level provided by the robotic system through performance indicators, which must somehow consider the measurement or estimation of human torque, which is a current challenge in rehabilitation robotics. This paper presents a control framework for partial assistance deployed within the Exo-H3 lower limb rehabilitation exoskeleton from the European Robotic Framework for bipedal locomotion benchmarking, H2020 EUROBENCH. Methods The control framework consists of three assistance controllers—angular trajectory tracking, interaction torque rejection, and a hybrid controller with two operation modes—; it also has direct control to compensate for motor gearbox frictions and the dynamics of the exoskeleton. The technical assessment of the control framework was carried out for each assistance controller on two healthy adult subjects, male and female, wearing the Exo-H3 exoskeleton and walking on a treadmill. The controller loop gain was modulated at four levels, such as 10%, 30%, 70% and 100%. Additionally, the BeStable testbed (Benchmarking System for Assessment of Balance Performance) and the Experience testbed (Benchmarking Exoskeleton-Assisted Gait Based on Users’ Objective Perspective and Experience) were used from the EUROBENCH framework. For all conditions and subjects, the assistance level was calculated from the total power and energy consumed by the exoskeleton and human and the proposed indicator, the exoskeleton assistance level; for the gait quality assessment, spatiotemporal gait parameters such as step length, step width, and step time were used. For each joint of the lower limbs, the absolute error integral of the tracking of the angular trajectory or interaction torque was calculated, as well as the total energy of the human and the exoskeleton, the percentage of the assistance level given by the exoskeleton and the dynamic time warping of the angular trajectory per joint. To assess the assistance level, the total human-exoskeleton energy and the exoskeleton assistance level were computed, and to assess the gait quality, the absolute error integral and the dynamic time warping were used for the trajectory pattern in the angular tracking controller and the spatiotemporal gait parameters. The performance of the three assistance controllers deployed within the Exo-H3 exoskeleton was assessed through one experimental protocol. Results The four designed controllers were deployed in the Exo-H3 exoskeleton, where the controller loop gain was modulated at four levels: 10%, 30%, 70%, and 100%. The performance indicators for each joint allow for assessing asymmetries in the lower limbs, as well as energy consumption and assistance level distribution. For the assistance controller with angular trajectory tracking, a loop gain of 100% attempts to fully track the angular reference, while a low level decreases the stiffness of the closed-loop system. Energy expenditure increased for one subject and decreased for the other. For both subjects, the angular trajectory tracking error decreases as the loop gain increases, and the error distribution in all joints is homogeneous; the minimum dynamic time warping between the reference and measured trajectories decreases as the loop gain in the controller increases, except for the hip joints; and the assistance level percentage per joint provided by the exoskeleton increases through the change in the controller's loop gains, except at 70% for Subject No. 1. Overall, the highest assistance level percentage is 40%, and the lowest is 12%. For the assistance hybrid controller with operation mode No. 1, a loop gain of 100% attempts to fully track the reference angular trajectory, indicating a high stiffness of the system, and a loop gain of 0% attempts to fully reject the interaction torque, indicating a low stiffness. Therefore, the exoskeleton provides partial assistance to the subject with loop gains between 0% and 100%. For both subjects, the energy expenditure during the change in the controller's loop gains is almost constant with small oscillations; the assistance level percentage provided by the exoskeleton for hip joints decreases, while that for knee joints increases and is almost constant for ankle joints. Overall, the assistance level percentage remains steady at approximately 45%, although it is based on the modulation of the controller's loop gain. Conclusions This work presents the design and assessment of a control framework for partial assistance deployed within the Exo-H3 lower limb exoskeleton. Four performance indicators and one experimental protocol are proposed to assess the controller’s performance based on assistance level and gait quality. According to the experimental results, the assistance controller with angular trajectory tracking achieved the best performance, with assistance level percentages between 12% and 40%.