Dynamic Optimization-Based Control of Dextrous Manipulation

We present a dynamic optimization-based control for dextrous manipulation with a multi-fingered hand. The formulation of the manipulation problem as an optimization problem enables computation of the best possible control with respect to user-defined criteria, trade-off between the many different and possibly conflicting equations the hand/object system must comply with, and easy removal or addition of such equations. In this paper, control of object motion and of contact forces is provided, as well as management of non-sliding contacts, motor constraints, joint limits and articular redundancy.