Manipulation allows a robot to alter its environment by performing mechanical work on it. Combined with locomotion, autonomous manipulation will be a key enabling capability in future robot systems for exploration, home care, military, industrial, and space applications. While most current robotic manipulation consists of simple grasp-and-carry operations, the spectrum of manipulation methods is much broader, including throwing, pushing, rolling, sliding, tapping, etc. Humans and animals make use of these methods on a daily basis, but robots do not. What is missing is a theoretical framework for combining and sequencing such manipulation methods for automatic planning and control of robotic manipulation.
This project is working toward autonomous manipulation by bridging the gap from individual robotic manipulation primitives to full manipulation sequence planning. The project is developing theoretical tools, a mechanics-based manipulation planning framework, feedback control algorithms, and an experimental manipulation testbed incorporating high-speed vision. The tools developed for autonomous planning and execution of manipulation sequences will allow future robot users to program robots at a much higher level of abstraction than currently possible. For example, instead of programming low-level control of motors, the user can simply say "place object A on object B." This will help grow the field of personal robotics, by allowing non-experts to interact with robots.
International partners in this work include leading laboratories in Italy and Japan. These collaborations provide unique access for graduate and undergraduate researchers to the world's leading experts in robotic manipulation.
