An improved human-robot interface by measurement of muscle stiffness

Abstract

Human contact with haptic devices introduces instabilities due to human operators' attemps to stiffen their arm to stabilize the system. Controllers often cannot measure arm stiffness and do not typically account for this. A method to effectively adjust the controller of a robotic force assist device to compensate for changes in operator arm stiffness was established. It was expected to achieve reduced oscillations and increased performance than one with fixed gains. The results could be used to design human-robot interfaces for force assisting devices. The compensating system used EMG signals to measure muscle activity, then estimated the stiffness of the human's arm. This was used to adjust the parameters of a haptic device's impedance controller based on a threshold. The system was then implemented on a small haptic device to study the effects with a human subjects. EMG signals were experimentally validated as an effective prediction of the stiffness of an operator's arm. The system was assessed in terms of performance and was found to provide improved stability and demonstrated the potential for increased performance. I. INTRODUCTIONHaptic devices require physical contact between the operator and the machine, using force feedback and creating a coupled system including the operator and the robot. This introduces inherent instabilities due to the typical response of human operators [1][2][3][4]. Instability in large haptically controlled force assisting devices which aid human operators in manipulating heavy or bulky loads can pose undesired risks to the load and the operator. Human operators often attempt to control oscillation by stiffening their arm,leading to a stiffer system with more instability. Controllers do not typically account for this, as the level of stiffness is often not directly measurable.This research will design a control scheme that adjusts to changes in the way an operator grips a haptic controller to attempt to correct for this. It will adjust the control parameters based on the overall stiffness of the operator's arm, which is expected to achieve higher performance in robotic devices than a fixed control system. This study will consider the effects of such a system on haptic force assisting devices. The control scheme must acquire some metric to estimate the operator's arm stiffness, and electromyogram (EMG) signals are expected to be suitable, as they provide information about the internal state of an operator's muscles. This will be validated by testing if EMG signals can be used to accurately indicate the stiffness level of a human arm. It should then be possible to design a control scheme with

Support Vector Machine classification of muscle cocontraction to improve physical human-robot interaction

“…For each antagonistic pair of muscles ( ), processed EMG values (f i , f j ) are scaled to respective maximum recorded EMG values (f i max , f j max ) to obtain normalized muscle activity (f i % , f j % ) values between 0 and 1. Cocontraction cc i, j of a muscle pair is determined by taking the minimum normalized muscle activity at each sample time [16,19,31], as shown in (8).…”

A Model for Operator Endpoint Stiffness Prediction during Physical Human-Robot Interaction

“…Eq. (7) displays the co-contraction index, cc i, j , a measure of muscle co-contraction determined by taking the minimum normalized muscle activity for a muscle pair at each sample time (Gallagher et al, 2012;Gribble et al, 2003;Thoroughman & Shadmehr, 1999). , ( ) = min( % ( ), % ( )) All measurements acquired (endpoint force, endpoint kinematics, and four EMG measurements) and directly computed variables during these experimental trials were subsequently used as simulation inputs, in order to compare and evaluate different prediction tools.…”