Learning interaction force model for endodontic shaping with support vector regression

Accurate estimation of interaction forces for endodontic shaping is fundamental to the interactive simulation of this operation. By applying new statistical learning techniques to this problem, this paper proposes a novel estimation method to acquire an optimized interaction force model to characterize input-output force mapping for endodontic shaping. We first present a novel robotic measurement system to acquire interaction forces in endodontic shaping and establish the needed training set. Then we propose a support vector regression model to learn the input-output force mapping for endodontic shaping. The regression model uses RBF kernel for training, and the optimized parameters of which are obtained by experiments. The learned model can convincingly estimate the interaction force resulting from endodontic shaping. And the effectiveness of the model has been evaluated by error measurement results