This research studies robot motion planning for tasks involving rolling and sliding contact with friction (generalized manipulation planning). The goal is to derive and use new results in multi-rigid-body mechanics and mathematical programming to develop techniques for the automatic generation of robust feedback control strategies. These strategies will guarantee correct execution of manipulation tasks with contact despite uncertainty in the parameters of the geometric, kinematic, and dynamic models, and sensing errors. When correct task execution cannot be guaranteed, the secondary goal is to specify a set of tighter, but widest possible, bounds on the uncertain parameters and sensing errors, which if satisfied, would guarantee correct execution. Application domains for the expected results include industrial assembly, grasp acquisition and dexterous manipulation, visual serving, medical assembly (e.g., hip replacement), virtual reality, and mechanical design. To test the practicality of the proposed approach and the validity of the underlying mathematical models, a CAD-based, manipulation planning system will be implemented. Automatically generated plans will be executed with real devices (i.e., an industrial robot and a planar dexterous manipulation testbed available at Texas A&M). In addition, real problems and associated data will be provided by colleagues from General Electric, Boeing, and the Amarillo National Resource Center for Plutonium. The work involves a collaboration with Profs. Nancy Amato and Jeffrey Trinkle of the Department of Computer Science at Texas A&M University.
