Force approximation and tactile sensor prediction for reactive grasping

Abstract

The sense of touch is an important component of dexterous manipulation. Multi-fingered robot hands with integrated tactile sensors are able to analyze and evaluate grasps. Therefore, predefined set-points are required to compare realized with nominal target values. Based on the comparison results, developed reactive grasping skills are able to optimize grasps. The question that arises here is whether and how tactile set-points as well as acting forces could be obtained. A preliminary analysis could provide information about important expectation values and grasp characteristics. Thus, force approximation and tactile sensor prediction are required. This paper describes an innovative approach of how to assign forces to received tactile sensor values as well as the design of a tactile sensor prediction. Additionally, potential fields of applications as well as some statistical analysis of received tactile sensor images are presented. It is shown, how to extend the designed algorithms to find reasonable grasp positions at the surface of complex objects.