Study of Gross Muscle Fatigue During Human-Robot Interactions

Abstract

This study explores the utility of Electromyogram (EMG) signals in the context of upper-limb exercises during human-robot interaction considering muscle fatigue of the participant. We hypothesise that the Electromyogram features from muscles and kinematic measurements from the robotic sensors can be used as indicators of fatigue and there is a potential to identify the muscle contribution during the activity where the Electromyogram data is correlated with the kinematic data. Electromyogram measurements were taken from four upper limb muscles of 10 healthy individuals. HapticMaster robot in active assisted mode together with a virtual environment was used to guide the participants for moving the robotic arm in a prescribed path in a horizontal plane consisting of four segments. The experiments were conducted until the participants reached a state of fatigue or until a defined maximum number of 6 trials were reached. Comparing the first and last trials indicated that the muscle fatigue had caused an increase in the average power and a decrease in the median frequency of EMG, which was more visible in Trapezius (TRP) and Anterior Deltoid (DLT) muscles in most of the analysed cases compared to Biceps Brachii (BB) and Triceps Brachii (TB) muscles. As the muscles came to a state of fatigue, the kinematic position also showed an increase in tracking error between the first and last trials. The ’near-thebody’ segment movements (S1 and S4 segments) were found to have less increase of tracking error compared to the ’away-frombody’ movements (S2 and S3 segments). A further analysis on this proved that the tracking error observed was mainly due to fatigue building up over the number of trials when performing ’away-from-body’ movements, and not a bi-product of perception errors. We identify that Deltoid and Trapezius muscles were fatigued more. These EMG fatigue indications can be mapped to kinematic indications of fatigue mainly in the segments S2 and S3, which required away from body movements because of the role of these two muscles in lifting the arm to the shoulder height in order to perform the activity. Our extracted features have shown the potential to identify the fatigued muscles as expected. The study also showed that the Electromyogram and kinematic features have a potential to be used to highlight the extent of muscle involvement.