¹CEA-LIST, Saclay, France
²BMBI, University of Technology of Compiègne, Compiègne, France
³Chemnitz University of Technology, Institute of Human Movement Science and Health, Department of Movement Science for Prevention and Rehabilitation, Chemnitz, Germany
⁴CIAMS, Université Paris-Saclay, Orsay, France
⁵M2S laboratory, INRIA, Université of Rennes 2, Rennes, France

Text

Introduction: Exoskeletons could nowadays be considered as a promising solution for assisting load carrying in industry, especially when dealing with active exoskeletons. The primary challenge inherent in this particular type of application lies in determining the mechanical actions to be applied to the subject’s anatomical segments in order to effectively relieve physical stress. Although the prediction of these forces may be computed through the use of electromyographic sensors (EMG) [1], [2], the way this assistance may be integrated into the exoskeleton control law, i.e. velocity, latency or amplitude, still needs to be investigated.

Methods: The objective of this study was to adapt an EMG-based control system of an upper limb exoskeleton [3] to a user based on specific individual characteristics. To this aim, a method has been designed to tune the parameters of control using objective criteria. More precisely, the user’s response time was used as an objective value to adapt the gain of the controller. The proposed approach was tested on 10 participants during a lifting task. Two different conditions have been designed to control the exoskeleton: (i) with a generic gain, and (ii) with a personalized gain. During the experiment, EMG signals were collected on five muscles to evaluate the efficiency of the conditions, and the user’s adaptation.

Results: Results showed a statistically significant reduction of mean muscle activity of the deltoid between the beginning and the end of each situation (28.6 ± 13.5% to 17.2 ± 7.3% of Relative Maximal Contraction for the generic gain, and from 24.9 ± 8.5% to 18 ± 6.8% of Relative Maximal Contraction for the personalized gain). When focusing on the first assisted movements, the personalized gain induced a mean activity of the deltoid significantly lower (29 ± 8% of Relative Maximal Contraction for the personalized gain, and 37.4 ± 9.5% of Relative Maximal Contraction for the generic gain). Subjective evaluation showed that the system with a personalized gain was perceived as more intuitive, and required less concentration when compared to the system with a generic gain.

Discussion and conclusion: These outcomes suggested that personalizing a control system of an exoskeleton by using user’s response time tends to improve the quality of the interaction, both in terms of muscular activity reduction and intuitiveness. These encouraging results open up new perspectives. In particular, we need to explore the possibility of implementing person-specific parameters in a routine procedure.

Literatur

[1] Fleischer C, Hommel G. A human-exoskeleton interface utilizing electromyography. IEEE Transactions on Robotics.2008;24(4):872–82.
[2] Peternel L, Noda T, Petrič T, Ude A, Morimoto J, Babič J. Adaptive Control of Exoskeleton Robots for Periodic Assistive Behaviours Based on EMG Feedback Minimisation. PLoS One. 2016 Feb 16;11(2):e0148942. DOI: 10.1371/journal.pone.0148942
[3] Treussart B, Geffard F, Vignais N, Marin F. Controlling an upper-limb exoskeleton by emg signal while carrying unknown load. In: 2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE;2020.p. 9107–13.