Low-dimensional robotic grasping: Eigengrasp subspaces and optimized underactuation

This thesis introduces new methods for enabling the effective use of highly dexterous robotic hands, interfacing with the upcoming generation of neurally controlled hand prostheses, and designing a new class of simple yet effective grasping devices based on underactuation and mechanical adaptation. These methods share a common goal: reducing the complexity that has traditionally been associated, at both computational and mechanical levels, with robotic grasping in unstructured environments.;In this thesis, we propose using low-dimensional posture subspaces for dexterous or anthropomorphic hands Human user studies have shown that most of the variance in hand posture for a wide range of grasping tasks is contained in relatively few dimensions. We extend these results to a range of robotic designs, and introduce the concept of eigengrasps as the bases of a low-dimensional, linear hand posture subspace. We then show that a grasp synthesis algorithm that optimizes hand posture in eigengrasp space is both computationally efficient and likely to yield stable grasps.;The emerging field of neurally controlled hand prosthetics faces a similar challenge when using dexterous hand models: bridging the gap between incomplete or noisy neural recordings and the complete set of variables needed to execute a grasping task. We propose using an automated grasp planning component as an interface, accepting real-time operator input and using it to assist in the synthesis of stable grasps. Computational rates needed for direct interaction can be achieved by combining operation in eigengrasp space with on-line operator input. Furthermore, the eigengrasp planning space can also act as an interaction space, allowing the operator to provide meaningful input for the hand posture using few channels of communication.;Algorithmic approaches to low-dimensional grasping can enable computationally effective algorithms and interaction models. Hardware implementations have the potential to reduce the mechanical complexity and construction costs of a hand design, using concepts such as underactuation and passive mechanical adaptation. Instead of complex run-time algorithms, hand models in this class use design-time analysis to improve performance over a spectrum of tasks. Along these directions, we present a set of analysis and optimization tools for the design of low-dimensional, underactuated hands. We focus on tendon-based mechanisms featuring adaptive joints and compliant fingertips, and show how a number of design parameters, such as tendon routes or joint stiffnesses, can be optimized to enable a wide range of stable grasps.;A key prerequisite for robot operation in human settings is versatility, which, in terms of autonomous grasping, translates into the ability to reliably acquire and interact with a wide range of objects. In an attempt to match the abilities of the most versatile end-effector known, the human hand, many anthropomorphic robotic models have been proposed, with the number of degrees of freedom starting to approach that of their human counterpart. However, these models have proven difficult to use in practice, as the high dimensionality of the posture space means that finding adequate grasps for a target object is often an intractable problem.;The ability to effect change on the environment through object acquisition (grasping) and manipulation has the potential to enable many robotic applications with high social impact, including effective neural prostheses, robots for house care or personal assistance, etc. We believe that the methods presented in this thesis represent a number of steps in this direction, advancing towards a proven solution for reliable autonomous grasping in human environments.