Human-like Modeling and Generation of Grasping Motion Using Multi-Objective Particle Swarm Optimization Approach Chiraz Walha, Member IEEE and Adel M. Alimi, Senior Member IEEE REGIM-Lab.: REsearch Groups in Intelligent Machines, University of Sfax, ENIS, BP 1173, Sfax, 3038, Tunisia {chiraz.walha, adel.alimi}@ieee.org Abstract- This paper presents a multi-objective particle swarm optimization (MOPSO) approach with focus to the application to human-like grasp planning. We deal with the problem of a five-fingered hand where we use kinematics to compute the con-tact point’s position with the object. Two methodology are formulated to tackle the grasp planning problem: forward and backward method. When the forward approach consider the hand point of view and generate directly the configuration sets, the backward approach evaluate the contact points on the object and the final solutions are converted to hand configurations using the inverse kinematics of the hand. The generation step of the forward method uses a Guided Random Generation (GRG) process which takes into account the limitation of joint angles of the human hand to generate appropriate configuration sets. Two fitness functions are used and Pareto front solutions are found through these approaches. Finally, a comparison between MOPSO and four other multi-objective approaches is presented and our approach outperformed the other approaches in terms of time run-ning and distribution of Pareto front solutions. I. INTRODUCTION A. Grasp Planning Problem Grasp Planning in its simplest form deals with the problem of where to place the fingertips on the surface of the object. In fact, it’s a resource consuming problem due to the high dimensionality of the configuration space. The main purpose of a grasp planner is to explore the dexterous manipulation space of the multi-fingered hand and find the best configuration of the fingers that enables a stable grasp. With the development of flexible and highly integrated dexterous robotic hands, the research results on grasp planning can be applied to create systems with autonomous grasping capabilities. Constraints due to target object geometry, environment, hand kinematics and a task description must all be considered when forming a configuration of the hand [1,2]. A hand-configuration is defined by the wrist position and orientation cou-pled with the angles of the fingers joints. The grasp planning process depends on some parameters like the number of fingers, the wrench space, the contact point model between the fingertips and the object (friction model), the grasp oriented task (power grasp or precise grasp), etc. However, finding the positions of the fingertips on the object can be really competitive since some criteria like the stability of the grip or the minimization of the friction had to be taken into account. Therefore, grasp planning is considered as an optimization problem. B. Related works In the last decades, several grasp planners have been developed under two main approaches [3]: empirical and analytical. The empirical or knowledge-based approach is motivated by the imitation of the human hands behavior like [4] who presented a matching algorithm to select appropriate grasps from a database based on the shape of the object. The analytic approach is based on the study of the robotic hand from a mechanical point of view. We can find the works of [5] who used a genetic approach to tackle this problem, [6] presented a heuristic approach to plan a precision grasp for 3D objects and [7] who presented an SVM approach involving a combination of numerical methods to recover parts of the grasp quality surface with any robotic hand under the simulator “Grasp it!” [8]. International Journal of Computer Science and Information Security (IJCSIS), Vol. 14, No. 8, August 2016 694 https://sites.google.com/site/ijcsis/ ISSN 1947-5500