Our goal is to allow planners to exploit the forces and contacts between objects so as to carry out complex manipulation tasks in the presence of uncertainty.

Many robotic manipulation systems are position-centric: the planner considers the position of objects and the robot, and during execution the arm controller maintains a prescribed position set point. For manipulation settings with requiring tasks like inserting a key, wiping a table, parts assembly, sliding, tool use, etc., position-based planning and control is not enough. Our goal is to allow planners to reason about forces and contacts between objects, and reason about the force-control and force-sensing capabilities of a robotic arm. On a robotic system, this requires the design of controllers and state estimators that use imperfect information from force, tactile, and position sensors

Research Areas