BEHAVIOR PREDICTION OF THE CURLING STONE USING THE DATA ASSIMILATION

Abstract

In this study, we proposed a model for predicting the movement of curling stone using the Discrete Element Method, and identified model parameters based on actual measurements of the movement. Assuming that there is a difference in the friction generated front and back side in the direction of travel, we proposed a model with five coefficients. Image analysis was performed from the video of throwing stones in practice, and time-series data of stone movements were acquired. The parameters were identified so that the sum of the residual squares of the time histories estimated by DEM and those obtained from the image analysis was minimized. As a result, we were able to reproduce the stone orbit that almost matched the measured data. Parameter identification was performed on the data of 88 cases, and it was shown that the variation of the obtained parameters is consistent with the sensitivity analysis using the Hessian matrix and Particle Filter. Furthermore, based on the identified parameters, the initial velocity and injection angle for placing the stone at the target were examined, and the sensitivities of the two parameters were quantitatively shown.