Variable admittance control with sEMG-based support for wearable wrist exoskeleton

Abstract

Introduction: Wrist function impairment is common after stroke and heavily impacts the execution of daily tasks. Robotic therapy, and more specifically wearable exoskeletons, have the potential to boost training dose in context-relevant scenarios, promote voluntary effort through motor intent detection, and mitigate the effect of gravity. Portable exoskeletons are often non-backdrivable and it is challenging to make their control safe, reactive and stable. Admittance control is often used in this case, however, this type of control can become unstable when the supported biological joint stiffens. Variable admittance control adapts its parameters dynamically to allow free motion and stabilize the human-robot interaction.

Methods: In this study, we implemented a variable admittance control scheme on a one degree of freedom wearable wrist exoskeleton. The damping parameter of the admittance scheme is adjusted in real-time to cope with instabilities and varying wrist stiffness. In addition to the admittance control scheme, sEMG- and gravity-based controllers were implemented, characterized and optimized on ten healthy participants and tested on six stroke survivors.

Results: The results show that (1) the variable admittance control scheme could stabilize the interaction but at the cost of a decrease in transparency, and (2) when coupled with the variable admittance controller the sEMG-based control enhanced wrist functionality of stroke survivors in the most extreme angular positions.

Discussion: Our variable admittance control scheme with sEMG- and gravity-based support was most beneficial for patients with higher levels of impairment by improving range of motion and promoting voluntary effort. Future work could combine both controllers to customize and fine tune the stability of the support to a wider range of impairment levels and types.

Keywords: gravity compensation; proprioceptive feedback; stroke rehabilitation; surface electromyography; variable admittance control; visuomotor task; wearables; wrist exoskeleton.

Figure 1

(A) The forearm module of the eWrist, where the motor and worm drive can be shifted up in order to uncouple the handle from the motor (see orange arrow). (B) The Myo gesture control armband from Thalmic Labs. (C) Illustration of the wrist angular position θ and the referentials used to compute F_grav, i.e., the earth referential ℜ₀ in green, the forearm module referential ℜ₁ in red, and the hand referential ℜ₂ in blue.

Figure 2

Block diagram of the variable admittance scheme and the two controllers. The elements within the purple dotted line form the sEMG-based controller, where the sEMG signal is normalized based on the participant's MVC and converted to a force F_emg via the gain G_emg. The elements within the blue dotted line form the gravity compensation controller that produces a force F_grav based on the exoskeleton's orientation and normalized via the gain G_grav. Subsequently, F_emg and F_grav are fed to the admittance controller. The elements within the green line are implemented in the Teensy microcontroller and run at 1 kHz, while the elements within the red line run on the host computer at 60 Hz.

Figure 3

(A) Picture and (B) schematic of the experimental setup where stiffnesses perceived by the eWrist can be modulated by moving the spring k along l. (C) Results from the evaluation where in (1) B_n was kept constant, and in (2) B_n was adjusted dynamically. In (3), the system was manually excited with high frequency and magnitude oscillations.

Figure 4

Experimental setup of the visuomotor task.

Figure 5

(A) Calibration window of the VMT where sEMG activity, force, wrist angle and IMU readouts are displayed in real-time. (B) VMT during vertical and (C) horizontal movements. (D) Left: calibration phase where in (i) MVC (indicated with a red bar) is measured on the two selected electrodes (in yellow) for extension and flexion separately, in (ii) low sEMG activity is required to assess W_hand, in (iii) active and passive (with the help of the experimenter) ROM is assessed, and in (iv) the gains are adjusted. Right: description of the two different sequences for healthy and stroke participants. *The order of the conditions was pseudo-randomized across participants.

Figure 6

The three phases of a successful trial. The plain curve is the normalized (to ROM_pas) angular trajectory $\widehat{\theta }$, the dotted line is the target level, and the gray horizontal bars are the targets zones and the home zone. A successful trial (i.e., target acquired) is composed of the movement initiation phase (in green), the rise phase (in blue), and the stabilization phase (in purple). A target is reached when the cursor enters the corresponding gray zone.

Figure 7

The boxplots show the median, interquartile range (IQR), and min./max. values of ten healthy participants for all factor permutations. The adaptive damping condition is shown in blue and the non-adaptive damping condition in pink. (A) Normalized integrated jerk (NIJ) and (B) interaction torque ${\widehat{T}}_{int}$. (C) Maximal angular velocity ${\stackrel{.}{\theta }}_{max}$ and (D) acceleration ${\ddot{\theta }}_{max}$.

Figure 8

Percentage of acquired targets in stroke participants in all conditions. The median across participants is shown.

Figure 9

Score comparison derived from questionnaires for stroke participants where the median is shown. (A) Scores from the questionnaire assessing the mechanical support. The average score over all aspects is 66.1 ± 12.8 for sEMG and 64.1 ± 12.9 for Gravity. (B) Scores from the RTLX questionnaire. The average workload score excluding Performance (Grier, 2015) is 40.2 ± 25.2 for sEMG, 39.0 ± 21.4 for Gravity, and 44.8 ± 23.2 for Transparent.