Self-Aligning Robot Gripper Jaw Design

Mike Tao Zhang and Ken Goldberg 
<August, 1999 ~ August, 2001> 
ALPHA Lab, University of California, Berkeley. 

Try the Design Applet

Note: For better performance of the Java applet and the movies, please use Internet Explorer NOT Netscape.


   Assembly lines often require grippers. It is possible in many cases to compensate for the difference in part orientation using a parallel-jaw gripper with appropriate jaw design. The idea is to arrange contacts on each gripper jaw so that the part is aligned as it is grasped and yet cannot jam. 


   We consider jaws based on a set of trapezoidal jaw modules that maximize contact between the gripper and the part at its desired final orientation. Furthermore, jaws are constrained to capture and rotate the part to its desired orientation and achieve a form-closure grasp. Given the n-sided 2D convex projection of an extruded polygonal part, we present an implemented O(n3 log n) algorithm to efficiently construct optimal jaw design. The algorithm combines toppling, jamming, non-liftoff, accessibility, and form-closure analysis. We also develop an O(n log n) algorithm to find a tolerance class for jaws (specified as maximum and minimum material conditions) given the optimal jaw design. 

   We introduce new geometric functions and a new data structure, the toppling graph, to represent the mechanical and geometric properties of jaw design. We implemented our jaw design algorithm in Java and the resulting jaw designs were verified by physical experiments. 


Grippers can be the most design-intensive components of an assembly system. Although grippers are widely used for automated manufacturing, assembly, and packing, the design of gripper jaws is usually ad-hoc and can be a major limiting factor in automated assembly. 

The part is initially in the resting pose. It is necessary to rotate the part to the desired final orientation for assembly. The idea is to design the gripper jaws so as to align the part as it is acquired. We divide the grasp into three phases: pushing, toppling and fixturing. Initially, one jaw makes contact with the part (the “pushing contact”). The jaw pushes the part along the work-surface until the part makes contact with the “toppling contact” on the second jaw.  At this point the part begins to rotate  (topple) from its initial orientation to the desired orientation. During toppling, the part is constrained by two contacts and the surface. When the part reaches its final fixtured orientation other jaw surfaces stop its motion.  The optimal jaw design guides the part through these phases, avoids premature toppling, jamming, and liftoff, and fixtures the part in form closure with maximal linear contact.

Java Implementation:

We implemented our jaw design algorithm in Java. Users can draw parts, define the COM and the friction coefficients, chose the initial and the final orientations; the applet computes, ranks, and displays the resulting gripper jaw design (Click the image below to try the applet). 


We also conducted a physical experiment using an AdeptOne industrial robot and a pneumatic parallel-jaw gripper (Mecanotron serial number: 101167) with jaws we designed. We added two air regulators (Wilkerson serial number: R08-01-F000) to adjust air pressure for the gripper, and also used utility compression spring (Century serial number: C-606) to slow down the gripper motion.

The first experiment we conducted was to design four-contact jaws for a small lever from a standard videotape (Fuji serial number: 7410161160). Its natural resting pose is as follows:

We have to insert a peg, which is on the video tape case, into the hole of the part for assembly (see the figure below). Therefore, rotation of the part in the vertical plane is required. 

We conducted a successful experiment in sequence 1~5 for 50 trials (Click the image below to see a movie, 5M).

In the second experiment, we verified both optimal jaw design and the jaw design with minimal linear contacts between the part and the jaws (Click the images below to see movies). 



  • T. Zhang, G. Smith and K. Goldberg. "Compensatory grasping with the parallel-jaw gripper," in Algorithmic and Computational Robotics: New Directions, edited by B. Donald, K. Lynch, and D. Rus, A K Peters, Ltd., 2001.
  • T. Zhang and K. Goldberg. "Gripper contacts for part alignment," IEEE Transaction on Robotics and Automation (under review). [PDF]
  • T. Zhang and K. Goldberg. "Self-aligning robot gripper jaw design," International Journal of Robotics Research (in process). 
  • T. Zhang and K. Goldberg. "Internet-based design of self-aligning robot gripper jaws," in IEEE International Conference on Robotics and Automation, Washington D.C., 2002 (under review).
  • T. Zhang, Lawrence Cheung, and  K. Goldberg. "Shape tolerance of robot gripper jaws," in IEEE/RSJ International Conference on Intelligent Robots and Systems, Wailea, HI , 2001. [PDF]
  • T. Zhang and K. Goldberg. "Design of gripper jaws based on trapezoidal modules," in IEEE International Conference on Robotics and Automation, Seoul, Korea, 2001. [PDF]
  • T. Zhang, G. Smith and K. Goldberg. "Compensatory grasping with the parallel-jaw gripper," in 4th International Workshop on Algorithmic Foundations of Robotics, Hanover, NH, 2000. 

Media Coverage:


    This paper grew out of practical experiments with a commercial assembly. We would like to thank Brian Carlisle from Adept Technology, Inc. and Randy Brost from Eastman Kodak Co. for suggesting these experiments and Kevin Lynch from Northwestern Univ. for his elegant toppling analysis. We would also like to thank Robert-Paul Berretty from UNC Chapel Hill, and A. Frank van der Stappen and Mark Overmars from Utrecht Univ. for their contributions to our thinking about the toppling function and Gordon Smith for suggesting that toppling could be applied to grasping. Thanks also to K. “Gopal” Gopalakrishnan and Yong Liu, and Mark Moll from Carnegie Mellon Univ., and Peng Song and Vijay Kumar from the Univ. of Pennsylvania for helpful discussions. We also thank the anonymous reviewers of an earlier version of this paper for their constructive feedback. This work is supported in part by the National Science Foundation under DMI-0010069, CDA-9726389 and Presidential Faculty Fellow Award IRI-9553197. Research funding is also provided by Adept Technology, Inc., Ford Motor Co., and 2000 California State MICRO Grant 00-032.