Abstract:
This study aimed to develop a knee-assistive device while walking. The research was separated into two sections: the gait support simulation in MATLAB and the prototype of the device. Data on body part positions and ground reaction force were collected from three adult Thai participants walking at a speed of 1.5 m/s to calculate knee moment. The simulation section provides support moments during walking using machine learning and artificial stiffness control strategy (MLASCS), composed of the kNN model and the instantaneous artificial stiffness per body mass (IASPB) equations. The MLASCS was used to determine the proper amount of support moment required to assist walking, and its validation via the recorded data showed that it could reduce the total effort by up to 63.4%. In the prototype section, the posterior-support device was designed using a 3D printing filament and tested for durability. The control system used an actuator replicated from an MIT mini-cheetah servo motor that commanded various parameters such as angular, angular velocity, angular stiffness, angular damping coefficient, and angular moment and provided feedback in the form of angular angular velocity and angular moment. Due to a significant increase in delay time when connecting the microcontroller to the device, the sets of the if-else function called a state classifier combined with the IASPB equations were selected as the control system instead of the MLASCS. Efficiency testing was conducted using electromyography (EMG) sensors, which revealed mixed results that the device was sometimes helpful and sometimes not helpful. These may be due to an imperfect gait cycle, motor command delays, and misalignment of the device, indicating that further data collection and validation with more samples is necessary to verify the device's usefulness.
