Towards more Natural, Intelligent, and Effective Robotic Grasp and Manipulations

点击翻译按钮获取中文摘要

To date, in controlled and accurately modeled environments such as factory floors, robots grasp massive objects and carry them with blurring speed and unerring precision. But, outside of such settings and in natural environment, every day manipulation tasks that we take for granted-even as simple as fetching a pen-can stump the most sophisticated robots. As robots increasingly make their way into functional roles in everyday human environments, we need them to be able to interact and work effectively with, for, and around people. The long-term vision of this research is to develop new and innovative solutions to empower robots to interact, through grasping, with their environment and humans more naturally, with speeds, dexterities, and degrees of autonomy that are comparable to their human counterparts. This vision is motivated by enormous practical values. At the same time, it entails difficult challenges related to the design, formulation of the grasp (grasp planning), maintaining the grasp and manipulating objects (grasp control), and the intersection of these topics with cognitive and functional structure of human grasping. By addressing some of these challenges through carefully selected short-term objectives, the proposed research program promises to deliver new solutions that pave the path towards enabling its long-term goal. Objective 1 focuses on the formulation of new Grasp Quality (GQ) measures applicable to a wider range of grasps. The formulation of GQ measures is an important open problem. A GQ measure is the main tool used for the entire process of grasp planning, grasp evaluating, and grasp control. Objective 2 focuses on developing new theoretical tools and enabling technologies to include active compliance (similar to human's muscle stiffening) in grasping. The new formulation based on active compliance will lead to more natural and robust grasp using algorithms that are substantially more powerful than passive approaches. Objective 3 develops new control algorithms that treat the control of the arm and the control of the hand as a unified problem, allowing for better coordination of the hand and arm and avoiding otherwise sub-optimal and unfeasible solutions. An important part of the proposed research program is the training of highly qualified personnel (HQP) both at graduate and undergraduate levels. The training received by the HQPs will provide them with in-depth knowledge and specialized expertise that ultimately lead to their competitive advantage in today's job market and their future professional success. By addressing important open problems in robotic grasping, the proposed research program will produce quality publications and other measurable impacts that are important to the Canadian robotic industry at large.