Robust Slippage Degree Estimation Based on Reference Update of Vision-Based Tactile Sensor

In this paper, we propose a new slippage degree estimation method of a vision-based tactile sensor. The advantage of the vision-based tactile sensor is simultaneous acquisition of the slippage degree, a multidimensional force and a moment. A CCD camera captures the sensor surface which has regularly arrayed dots. The slippage degree on the sensor surface is determined by the stick ratio obtained from the displacements of dots in the captured images. In our previous work [1], the stick ratio was successfully applied to avoid slippage or the shape deformation of grasped objects. Based on the adaptive selection of the reference image and the compensation of the dot displacement, the proposed method in this study extends use of the previously developed algorithm, to dynamically complex but general situations as follows. First, the contact surface deforms after the macroscopic slippage or the slippage direction is changed. Second, the contact surface rotates with an applied moment. Third, the captured image is locally zoomed with a significant change of a grip force. A heuristic weighted average method is also proposed to decrease each dot's variation in the captured image. Usefulness of the proposed method is confirmed through experimental results.