A Model-based Strategy for Mapping Human Grasps to Robotic Hands Using Synergies Fanny Ficuciello 1 , Gianluca Palli 2 , Claudio Melchiorri 2 and Bruno Siciliano 1 Abstract— The aim of this paper is to derive the synergies subspace of an anthropomorphic robotic hand using the human hand as a master. A set of grasping postures performed by five subjects in grasping commonly used objects has been mapped to a robotic hand assuming its own kinematics as a simplified model of the human hand. Using an RGB camera and depth sensor for 3D motion capture, the human hand palm pose and fingertip positions have been measured for the reference set of grasping. From the measured fingertip positions a closed-loop inverse kinematics algorithm has been applied to reproduce the joint space configuration of the robotic hand relying on its kinematics, scaled using the human and robotic fingers length ratio. Once the set of grasping has been mapped on the robotic hand, the synergies subspace has been computed applying principal component analysis on the joint configurations. The obtained subspace is tested with experiments on the DEXMART Hand by performing reach to grasp actions on selected objects using the first three predominant synergies. The analysis of these synergies and a comparison with the results on the human hand available in the literature are performed by means of graphical and numerical tools. I. I NTRODUCTION The use of the principal components, also called postural synergies, holds great potential for robotic hands control, implying a substantial reduction of the grasp synthesis prob- lem dimension. Transferring human hand motion to a robotic hand is a quite challenging problem due to the complexity and variety of hand kinematics and the dissimilarity with the robotic hand. Indeed, in order to obtain a thorough human hand posture estimation, a reliable kinematic hand model and high-accurate motion tracking instrumentation are required. A synergies mapping from the human hand to the robotic hand has been addressed in [1]. The proposed mapping strategy between the synergies of a paradigmatic human hand and a robotic hand is carried out in the task space and it is based on the use of a virtual sphere. In [2] three synergies have been extracted from data on human grasping experiments and mapped to a robotic hand. Thus, a neural network with the features of the objects and the coefficients of the synergies has been trained and employed to control robot grasping. In [3], postural synergies of the UB Hand IV, the first prototype of the DEXMART Hand, 1 Dipartimento di Ingegneria Elettrica e Tecnologie dell’Informazione, Universit` a degli Studi di Napoli Federico II, 80125 Napoli, Italy, email: {fanny.ficuciello, bruno.siciliano}@unina.it. 2 Dipartimento di Ingegneria dell’Energia Elettrica e dell’Informazione “Guglielmo Marconi”, Alma Mater Studiorum Universit` a di Bologna, 40136 Bologna, Italy, email: {gianluca.palli, claudio.melchiorri}@unibo.it. This research has been partially funded by the EC Seventh Framework Programme (FP7) under grant agreement no. 287513 for the IP SAPHARI (Safe and Autonomous Physical Human-Aware Robot Interaction). Fig. 1. Diagram of the methodological approach. have been used for human-inspired control solutions. In [4], neural networks have been integrated to allow synergy-based grasp planning relying only on object geometric features and task requirements; the experiments were conducted on the last prototype of the DEXMART Hand. In this work, we propose a new model-based method to map grasps from the human to the DEXMART Hand [5]. The core idea of the mapping method is to dramatically simplify the detection of the human hand posture by measuring strategic points using a low cost camera and using the robotic hand kinematics as a paradigmatic model of the human hand for inverse kinematic computation. The Kinect sensor is used for 3D human hand fingertips detection and a Closed-Loop Inverse Kinematics Algorithm (CLIK) [6] is used for human grasp mapping and is based on the robotic hands kinematics linearly scaled according to the subjects’ hand dimensions. The paper is organized as follows: Section II describes the DEXMART Hand kinematics. Section III explains the technical approach on the whole, from human hand detection to synergies derivation. Section IV illustrates the mapping procedure based on measures acquired with the Kinect. In Section V the computation of synergies subspace and the experimental results obtained using an open-loop control strategy based on the first three predominant synergies are reported. In Section VI the first three predominant synergies of the DEXMART Hand are analyzed by means of graphical and numerical tools. Evaluation of the method efficiency is made on the basis of the analysis performed in [7] on the human hand and taking into account the results of the mapping obtained on the UB Hand IV using the same reference postures. Finally, Section VII provides the conclusion. 2013 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM) Wollongong, Australia, July 9-12, 2013 978-1-4673-5320-5/13/$31.00 ©2013 IEEE 1737