Grasp and Manipulation Analysis
The theory of grasp and manipulation analysis has many established results that are foundational for the work done today. Still, our current analysis tools have not yet enabled tasks requiring high level of dexterity, such as object acquisition from dense clutter, dexterous manipulation of generic objects, etc. Are there key tools still missing from our analysis toolbox? We are looking for the computational tools that will enable applications such as general grasp analysis for dexterous hands, manipulation and grasp planning, hand analysis, etc.
Project highlights (in chronological order):
We investigate the distinction between active force generation and passive force resistance in grasping. This distinction is missing from some of the most commonly user grasp analysis tools; still, we believe it must be considered for most types of hands used in practice. [Video below shows that kinematically identical grasps can lead to dramatically different results depending on how contact forces respond to commanded actuation forces. In the case of this underactuated gripper, loading joints differently (by different tendons) leads to very different contact forces, in turn leading to different ability to resist disturbances. Our framework correctly captures this difference.] [Publications: T-ASE 2018, WAFR2016, IROS 2016 Workshop]
We then introduced the first algorithms capable of solving this problem with strong guarantees, and in a computationally efficient fashion. The first algorithm was for 2D grasps, taking advantage that the number of possible slip directions at any contact is finite [RSS 2018]. In 3D, where the same problem is non-convex, convex relaxation method, we introduced a convex relaxation method along with an algorithm that successively refines this relaxation locally in order to obtain solutions to arbitrary accuracy efficiently. Our resulting algorithm can determine if a grasp is passively stable, solve for equilibrium contact forces and compute optimal actuator commands for stability [IEEE T-RO 2020].