Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 02, 2018 - June 05, 2018
Tactile information is important when manipulating objects with robot hands. We propose a random-dot type optical tactile sensor using a camera to acquire tactile information. In the random dot type, a large number of grains are embedded in the soft gel of the contact part. This sensor acquires the displacement of dots with a camera image reflecting deformation of gel due to object contact. In order to extract various tactile information from the image, we analyze the image by using a deep neural network. In this study, we evaluate the performance of shape discrimination of the contact objects and force estimation.
