NRI: Multi-Digit Coordination by Compliant Connections in an Anthropomorphic Hand

University Of Tulsa, Tulsa OK

Investigators

Abstract

Robust, soft, lightweight, dexterous hands will be a crucial component of the next generation of robots that physically interact with human co-workers. To date, under-actuated robot hands (where one motor moves multiple joints) that grasp objects reliably have been created largely by designer ingenuity using "tricks of the trade" without fully capitalizing on the wealth of mathematical grasping and manipulation theory. Moreover, anatomical studies of the human hand and robotic hands that do useful work have largely evolved separately, leading to similar conclusions expressed in a different way. This collaborative effort between engineering and medical faculty will help bridge these gaps. Engineering graduate students and medical residents will interact regularly; this will serve to translate knowledge and experience across the boundaries of these two disciplines, thereby fostering the ability to solve complex, multi-faceted bioengineering problems. A key contribution of this project will be a theoretical description of an elastic drive train that maps individual actuator motions to key movements in tasks of daily living. The robotic hand produced as another outcome of this work will serve as a robust platform for future experimentation (for example, in studies on gesturing, communication for the hearing impaired, and tactile exploration), and it will also be useful for evaluating and visualizing the effect of potential surgical interventions ex vivo. An important educational and research tool in the PI's institution, this hand will be used to educate and interest university and K-12 students in robotics. The PI team will coordinate with student groups and Native American community organizations, to recruit qualified underrepresented minorities to participate in the project. Moving from single- to multi-actuator multi-fingered dexterous robot hand design is not straightforward and has frustrated researchers for several years; it will require a principled formal treatment. In single-actuator hands, adding a spring to the drive train improves the hand; the exact spring characteristics are not critical. In the general case of multi-actuator compliant hands simply adding a spring is not enough; a multidimensional spring of specific characteristic is needed. Concepts from multiport circuit theory, mechanics of materials, linear algebra and elements of general multi-fingered grasping theory will be used to understand the behavior of such a multi-dimensional spring. The PI's approach is to break this problem into sub-problems, greatly simplifying the analysis. There is also the matter of which actions an under-actuated hand should be able to perform; the PI argues that complete finger individuation is not necessary, rather coupled finger movements should correspond to those used by humans to perform tasks. Anatomical analysis of both normal humans and those that have undergone surgical intervention will be used to prioritize basic human motion primitives. Combined with the novel compliant grasping theory, under-actuated hand synthesis will be formulated as a parameter fitting problem. Detailed evaluation of the effects of surgical intervention on the human hand, the study of human anatomy, multi-port circuit models, and consideration of compliant mechanisms and the fundamental subspaces of linear algebra will all be blended. Taken together, this will result in a method to synthesize an elastic drive train that maps actuator contributions to useful, coordinated motions of the fingers. Implications of the compliant actuation theory will be examined in the mechanism synthesis and grasp stability contexts, and these will be evaluated experimentally. The investigators will pursue this line of reasoning all the way to implementation in a functioning anthropomorphic hand capable of performing the most common tasks of everyday life.