Motion Planning of Robot Fingertips for Graspless Manipulation Yusuke MAEDA Dept. of Prec. Eng., School of Eng., The University of Tokyo 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656 JAPAN maeda@prince.pe.u-tokyo.ac.jp Tomohisa NAKAMURA NTT Data Corporation Tamio ARAI Dept. of Prec. Eng., School of Eng., The University of Tokyo 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656 JAPAN arai@prince.pe.u-tokyo.ac.jp Abstract— In this paper, we present a method of motion planning of multiple robot fingertips for graspless (or nonpre- hensile) manipulation. The method can automatically generate various graspless operations, including pushing and tumbling. By considering whether each robot finger should be position- or force-controlled, we can obtain robust manipulation plans against external disturbances. Some examples of planned graspless ma- nipulation of a cuboid by two robot fingers are presented. We also show an experimental result of execution of planned graspless manipulation by a robot with a multi-fingered hand. I. I NTRODUCTION Manipulation without grasping is referred to as graspless manipulation [1] or nonprehensile manipulation [2]. In this paper, we study graspless manipulation where the manipulated object is supported not only by robot fingers but also by the en- vironment; it includes pushing, sliding and tumbling (Fig. 1). Graspless manipulation brings the following advantages to robots: • Manipulation without supporting all the weight of the object • Manipulation with simple mechanisms • Manipulation when grasping is impossible Thus graspless manipulation is important as a complement of conventional pick-and-place to enhance the dexterity of robots. Planning of robot motions to move an object from an initial configuration to a goal is a fundamental problem in robotic manipulation. However, robot motion planning for graspless manipulation is much more difficult than that for pick-and-place [3]. In pick-and-place operation, once an object is grasped, the correspondence of its motion to the robot motion is trivial; therefore manipulation planning is reduced sliding tumbling pivoting pushing Fig. 1. Graspless Manipulation to a geometrical collision avoidance. On the other hand, the correspondence in graspless manipulation is nontrivial; thus manipulation planning requires mechanical analysis for consideration of the effect of gravity, contact forces, and so on. Moreover, graspless manipulation may be irreversible: For example, a robot can push an object but may not be able to pull it back. Therefore, most of related studies deal with planning of manipulation with a specific operation such as pushing (e.g., [4], [5]). The authors proposed a planning method for general grasp- less manipulation by multiple robot fingertips [3]. However, the method can deal with only planar graspless manipulation. In this paper we extend our previous method. The main improvements are as follows: • Now it can deal with spatial graspless manipulation such as pushing and tumbling of a polyhedron by multiple robot fingertips. • Control modes of robot fingers (force control/position control) are considered explicitly for realistic manipula- tion. • A new stability measure [6] is adopted in planning to generate robuster graspless manipulation. • A* algorithm [7] is used to accelerate planning. This paper is organized as follows: Section II introduces a model of graspless manipulation. Section III describes a method to determine appropriate finger control modes (force control or position control) at an instant in graspless manipu- lation based on [8]. Section IV proposes a method of motion planning of robot fingertips for graspless manipulation. In Section V, some examples of planned graspless manipulation including pushing and tumbling are presented. We also show an experimental result of execution of planned manipulation by a robot with a multi-fingered hand. Finally, this paper is concluded in Section VI. II. MODEL OF GRASPLESS MANIPULATION A. Assumptions In this paper, for graspless manipulation by multiple robot fingertips, we make the following assumptions: 1) The manipulated object, robot fingertips, and the envi- ronment are rigid. Proceedings of the 2004 IEEE International Conference on Robotics & Automation New Orleans, LA • April 2004 0-7803-8232-3/04/$17.00 ©2004 IEEE 2951