This National Robotics Initiative project will promote the progress of science and advance the national health, prosperity and welfare by discovering foundational scientific knowledge needed to allow robots to achieve human-like dexterity in constrained manipulation tasks. Current robots are better than people at executing desired motions in free space, but they are not nearly as adept at performing tasks in which motion is constrained in some way. Examples of such tasks in everyday life include peeling potatoes, opening a bottle, closing a container, cleaning a surface, or assembling furniture. Robots have great difficulty in performing these types of tasks because inevitable uncertainties in object exact locations often cause robot commanded positions to conflict with physical constraints, resulting in excessive contact forces and task failure. This award supports fundamental research into improving robot dexterity by providing a robot system the human-like ability to continuously adjust its inherent mechanical behavior as a task progresses so that it can be compliant in directions that are constrained and to be stiff in directions for which the commanded motion is needed to complete the task. Human-like dexterous manipulation is needed for robots to be widely used in senior living-assistance, agriculture, construction, space exploration, healthcare, nuclear remediation, and manufacturing. The robustness of the approach will be demonstrated in a manufacturing assembly testbed application. Therefore, results from this research will provide broad benefit to society and the U.S. economy. The technologies developed will be in the areas of mechanical engineering (kinematics, dynamics, and control) and computer science and will engage and educate students, including members of underrepresented groups, in these areas of national need.
To obtain improved dexterity, multiple collaborating serial manipulators will be used to achieve a desired time-varying compliance, one specifically optimized for a given task so that it is properly executed despite uncertainties in task geometry and associated physical constraints. The compliance will be passively realized using variable stiffness actuators in each joint of the multi-manipulator system. With this optimized modulated passive compliance approach, the inherent mechanical behavior of the robot system is guaranteed to regulate contact force to not exceed specified limits and to move as desired to attain task objectives. The three specific aims of this research project are: 1) identify necessary and sufficient conditions and general synthesis procedures to achieve any specified passive compliance for different multi-serial manipulator topologies, 2) identify procedures for continuously and simultaneously attaining the desired object position and passive elastic behavior as an object is moved by redundant collaborating robots for different multi-serial manipulator topologies, and 3) demonstrate customizable dexterous manipulation using a three-finger planar robot hand with modulated passive compliance (using novel antagonistic tendon-driven variable stiffness actuators) performing a variety of different planar assembly tasks quickly and reliably.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
