Detecting Feature of Haptic Interaction Based on Distributed Tactile Sensor Network on Whole Body

Abstract

To extract information about users contacting robots physically, the distribution density of tactile sensor elements, the sampling rate, and the resolution all must be high, increasing the volume of tactile information. In the self-organized skin sensor network we propose for dealing with a large number of tactile sensors embedded throughout a humanoid robot, each network node having a processing unit is connected to tactile sensor elements and other nodes. By processing tactile information in the network based on the situation, individual nodes process and reduce information rapidly in high sampling. They also secure information transmission routes to the host PC using a data transmission protocol for self-organizing sensor networks. In this paper, we verify effectiveness of our proposal through sensor network emulation and basic experiments in spatiotemporal calculation of tactile information using prototype hardware. As an emulation result of the self-organized sensor network, routes to the host PC are secured at each node, and a tree-like network is constructed recursively with the node as a root. As the basic experiments, we describe an edge detection as data processing and extraction for haptic interaction. In conclusion, local information processing is effective for detecting features of haptic interaction.