GraspIt!: A versatile simulator for robotic grasping.

摘要

Grasping in an industrial setting is generally accomplished with low complexity grippers which offer a high degree of reliability within a structured environment. But, as robots are moved out of the factory into more unstructured settings, they will need more flexible devices to grasp or manipulate objects. To fill this need, articulated hands are being designed with more and more degrees of freedom, but this additional flexibility comes at a price of additional complexity in the design process and in choosing a grasp from the large number of possibilities. Mechanical grasping theory, which has been thoroughly studied in the last 20 years, provides objective methods for evaluating grasps, but the field has lacked a tool that can apply this analysis to arbitrary grasps of objects using different robotic hands.; This thesis presents a grasping simulator that can aid both robotic hand designers and those planning grasping tasks. It provides a user with an interactive environment where he or she can manipulate the degrees of freedom of any given hand model to form grasps. As contact between the links of the hand and an object occur, the simulator analyzes the new grasp on the fly and provides the user with instant feedback concerning the quality of the grasp. This information includes not only numerical quality measures, but also an indication of the grasp's weak point as well as projections of the 6-dimensional space of forces and torques that can be applied by the grasp.; In addition to its visualization capabilities, the simulator has facilities for modeling a complete robotic work cell, which is necessary when planning an actual grasping task. When combined with a real-time, model-based vision system to align the pose of a simulated object with the pose of an actual object in the workspace, the simulator can be used to plan, execute, and monitor a complete grasping task. Examples of this process are presented and serve to demonstrate this application of the simulator.; Finally, the addition of a module to simulate object dynamics further increases the realism of the simulation. With it, a user can examine the grasp formation process to ensure that a desired grasp can be achieved. Examples in this section demonstrate the benefit of going beyond a spatial analysis and including the temporal dimension.