Gait environments recognition using Gaussian process regression model-based CoP trajectory for wearable robot applications

Abstract

This paper proposes a new method for estimating the foot plantar center of pressure (CoP) trajectory using an insole sensor and Gaussian process regression (GPR) model. The proposed method is designed to interpolate the points between the sensor array, making it robust for different gait environments, such as plain, uphill slope, downhill slope, and upstairs. Three subjects with different foot sizes were attended to evaluate the performance of the method, and the results were compared and analyzed. This study found that the CoP trajectory is estimated differently depending on the gait environments, and the correlations between the estimation results were analyzed using statistics obtained through box plot and RMSE (inverse correlation) values. The first subject showed the largest difference according to gait environments, while the third subject showed the smallest difference. Additionally, toe angle is defined to compare the differences from the expected CoP trajectories according to the walking environments. The uphill slope and upstairs showed the outside direction of the toe, while the downhill slope showed the inside direction of the toe. The study also found that non-plain (or non-flat) environments such as slopes and staircases were able to be recognized through comparison with the plain environment. The difference between the slopes and the stairs could be distinguished according to the heel strike duration. Finally, the developed algorithm will be applied to the wearable robotic system under development. Overall, the proposed method shows potential for robustly estimating the foot CoP trajectory in different gait environments and could have practical applications in wearable robotic systems.