Fingertip Contact Detection for a Multi-fingered Under-actuated Robotic Hand using Density-based Clustering Method

Robotic Hand using Density-based Clustering Method

Abstract

Detecting the contact of Multi-fingered robotic hand’s fingertip and an object plays an important role in grasping the object stably. However, there remained considerable difficulties due to non-ideal external factors, in which for an under-actuated finger with force sensor located at the link far from the fingertip, each contact position cannot be determined using only the kinematic model and tactile information. In this paper, the Density-based clustering method to the force sensor and position encoder data is introduced to realize real-time contact detection. In our approach, data of the finger’s joint sensor during operation when not in contact with object is collected and analyzed using Density-based clustering algorithm. Learned information and collected data are then used to perform real-time contact detection.