Shape Classification in Continuous Rotation Manipulation by Universal Robot Hand

Abstract

We propose a shape classification method in continuous rotation manipulation by a multi-fingered robot hand. Our robot hand has five fingers equipped with tactile sensors. Each tactile sensor can measure a pressure distribution once every 10 ms while the robot hand rotates an object continuously. Our proposal classification method consists of the following processes: A kurtosis is calculated from each pressure distribution, and it quantifies the shape of the current contact surface. By rotating an object and measuring a time-series pressure distribution, the hand obtains a time-series kurtosis. An evaluated value is calculated from the time-series kurtosis and reference patterns using continuous dynamic programming (CDP) matching scheme. The contact shape is classified whether the evaluated value is lower than a given threshold. Experiments demonstrate the effectiveness of our proposal shape classification.