인공신경망 기반 무릎 관절 각도 추정 시스템 개발 (Development of Knee Joint Angle Estimation System based on Artificial Neural Network)

In order to estimate the knee joint angle to control the exoskeleton robot, many sensors and equations require many parameters.
Since this is unfavorable for the commercialization of exoskeleton robots, a new MLP method was attempted to compensate for this.
In this study, a study was conducted to estimate the knee joint angle based on plantar pressure data to identify the wearer's movement intention using a walking aid exoskeleton robot. This is a pHRI-based system used to determine the wearer's intention in the exoskeleton system, and the angle of the wearer's knee joint was estimated through the planter pressure data. A device capable of simultaneously measuring plantar pressure and knee joint angle was designed, and data were obtained by conducting a walking experiment with 10 adult male subjects on a treadmill. An artificial neural network was trained through the obtained data, and a study was conducted to estimate the knee joint angle using the planter pressure data through the trained artificial neural network.
The MLP model trained on 10 subjects each had an R2 Score and Adjusted R2 Score close to 1. It showed R2 Score and showed high performance.

In order to estimate the knee joint angle to control the exoskeleton robot, many sensors and equations require many parameters.
Since this is unfavorable for the commercialization of exoskeleton robots, a new MLP method was attempted to compensate for this.
In this study, a study was conducted to estimate the knee joint angle based on plantar pressure data to identify the wearer's movement intention using a walking aid exoskeleton robot. This is a pHRI-based system used to determine the wearer's intention in the exoskeleton system, and the angle of the wearer's knee joint was estimated through the planter pressure data. A device capable of simultaneously measuring plantar pressure and knee joint angle was designed, and data were obtained by conducting a walking experiment with 10 adult male subjects on a treadmill. An artificial neural network was trained through the obtained data, and a study was conducted to estimate the knee joint angle using the planter pressure data through the trained artificial neural network.
The MLP model trained on 10 subjects each had an R2 Score and Adjusted R2 Score close to 1. It showed R2 Score and showed high performance.