1. Grasping Curved Objects through Rolling
In the recent one and a half years I have been studying how to grasp an object with curved boundary by rolling fingers on it. Each finger is equipped with a tactile sensor capable of recording any instantaenous contact with the object. This task can be divided into two phases: finger localization and grasp achievement. In the first phase, the objective is to estimate the finger placement on the object from tactile data. In the second phase, the objective is to control the rolling of the fingers to locations on the object where a force-closure grasp can be formed.
We assume that the object is free to slide in the plane as the fingers are rolling on its boundary. In 2-D, two fingers are needed to cancel out the unknown angular velocity of the object, in the sense of comparing their “feelings” against each other. Here the “feeling” refers to the tangent rotation angle, i.e., total curvature, on the object boundary as determined by individual contact movement (see the figure on the left). The difference between the “feelings” on the two fingers can be calculated from the tactile data. More formally speaking, contact kinematics reveal a relationship between the amount of finger rotations and the total curvatures of the boundary segments of the fingers and the object respectively traversed by the two contact points during the same period of rolling. Such relationship makes it possible to localize both fingers relative to the object from a few pairs of simulataneously taken finger contacts at different time instants. Currently, a least-squares forumlation is set up to solve for the finger locations. And after localization, a simple open-loop strategy is used to control the continual rolling of the fingers to reach locations where a grasp can be formed. The weakness of the current finger localization algorithm is that it is local. I have been working on experiments with an Adept Cobra 600 robot.
- Yan-Bin Jia. Grasping Curved Objects through Rolling. In Proceedings of the IEEE International Conference on Robotics and Automation, pp. 377-382, San Francisco, CA, April 2000.
2. Computing Optimal Planar Grasps
Previously I also worked on optimal grasping. In mechanis and robotics community the quality of a grasp can often be measured as the magnitude within which any external wrench is resistible by “unit grasp force”. The figure on the right shows a four-finger force-closure grasp on a 5-gon. I presented a numerical algorithm to compute the optimal force-closure grasp on a simple polygon, given contact forces of unit total magnitude. Forces are compared with torques over the radius of gyration of the polygon. I also addressed a grasp optimality criterion for resisting an adversary finger located possibly anywhere on the polygon boundary. The disparity between these two grasp optimality criteria were demonstrated by simulation with results advocating that grasps should be measured task-dependently. The work assumes non-frictional contacts.
For more please see the following paper:
- Yan-Bin Jia. On computing optimal planar grasps. In Proceedings of the 1995 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 3:427-434, Pittsburgh, PA, August 1995.