The Manipulation Laboratory at Carnegie Mellon University investigates fundamental modes of manipulation under uncertainty. Our goal is to produce robots that can perform a variety of tasks in the physical world, ranging from industrial assembly to everyday chores. Examples include prepositioning parts for camcorder assembly to sorting papers on a desktop.
Our key contribution has been to elucidate and harness natural sources of information. Often, when asked to name sources of information researchers think of simple sensors such as cameras, light detectors, and strain gauges. They forget that action is as much a source of information as it is a source of uncertainty. Dropping a book on a desk is often the best way to know that the book is on the table. More complicated mechanical information appears in dynamic settings. Pushing on a coffee cup with a sensating finger provides a combination of mechanical and sensory information that allows the robot to determine the cup’s shape and pose.
Many techniques such as these were first developed in the Manipulation Laboratory in the context of hands and arms interacting with objects in the world. They are now commonplace in other domains, such as mobile robotics.
Software automation has become a natural part of human life. In the future this automation will extend to physical devices. Today we can place orders for goods over the internet, but humans still pack and ship these orders. In the future, we will be able to go grocery shopping without human intervention. Today, we can call up cameras to view scenery in Hawaii or New York. In the future we will be able to sample the plants and rocks we see at such distances. The Manipulation Laboratory is contributing with fundamental research into the nature of action and sensing and their natural connections.
Is it possible to build robots which take the desktop as their task domain?
The goal of this project is to show how robots can gain access to difficult-to-reach places by tak...
"Extrinsic Dexterity" is a way to get dexterous manipulation with a very simple hand, by coordinat...
Designing simple grippers for autonomous general purpose manipulation.