Food Feature Learning for Knife Cutting Operation of Cooking Robot with Parametirc Bias

Abstract

The knife cutting operation of food in cooking is a task that requires dealing with nonlinear dynamics for a variety of ingredients with different types and individual differences. In this study, Deep Predictive Model with Parametric Bias (DPMPB) was used to learn state transition models and foodstuff features. We conducted an experiment to learn foodstuff features for 10 kinds of foodstuffs in the cutting operation of the robot, and confirmed that the foodstuff features were acquired as Parametric Bias without any supervision.