Variable stiffness ankle exoskeleton robot Estimation of ankle joint stiffness value over EMG signals Real-time implementation of EMG based ankle stiffness estimationFigure A. Impedance control of VS-AnkleExo and position control diagram of the stiffness adjustment mechanism to follow the estimated ankle stiffness valuePurpose: The purpose of this study is to estimate the ankle joint stiffness with using the ankle musculoskeletal model parameters obtained via EMG signals and to show the real-time implementation of these predicted stiffness values on an ankle exoskeleton robot (VS-AnkleExo).
Theory and Methods:In this study, a musculoskeletal model approach was used to describe the behavior of the ankle joint. The model consists of a joint driven by two muscles that provide plantar-flexion and dorsi-flexion movements of the ankle. In order to obtain the muscle forces that provide the necessary movements of the ankle, it was used Mykin muscle model. The parameters in the Mykin model were found by the help of measured torque and EMG data obtained with different feature extraction methods such as Root Mean Square, Mean Absolute Value, Average Amplitude Change, Difference in Absolute Standard Deviation Value, v-Order, Log Detector, Zero Crossing and Slope Sign Change. The Mykin model parameters were changed according to the feature extraction methods. Therefore, a verification experiment was carried out to decide which signal processing method is the best, and it was decided that Slope Sign Change method is the most appropriate one. Then, the stiffness estimation of the ankle joint was performed by using the biomechanical parameters found with this method. Finally, the estimated stiffness value was sent to actuation unit of AnkleExo in real-time, and the process of applying the force feedback impedance control algorithm was carried out.
Results:The test results obtained with the Slope Sign Change method show that the estimated ankle joint stiffness value varies continuously between 200 Nm/rad-700 Nm/rad. Furthermore, it has been observed that the interaction torque between the user and the exoskeleton robot can be kept about 25% lower levels while VS-AnkleExo follows the ankle position by constantly changing the stiffness value in real-time.
Conclusion:The estimated ankle stiffness values revealed in this study match with the ankle stiffness values found in the literature. Besides, it is seen that lower interaction torques occurred between the user and the robot during imitation of the stiffness values of the ankle in a real-time implementation of an impedance control.