Abstract: One of the main weaknesses of intelligent robotic systems today is their inability to reason generally about contact. This prevents them from planning and performing grasping and dexterous manipulation in unstructured environments such as homes. In manufacturing settings, robotic workcells are typically highly structured with tight tolerances, so much so, that these “flexible” workcells have little flexibility and high implementation costs. These two issues alone present significant drags on the growth of the personal robotics market and penetration of robotic systems in the manufacturing of personalized and low-volume products.
Results in time-stepping methods for multibody systems, optimization techniques, and supporting software have presented roboticists (and their robots) with some of the technical components needed to support simulation-based planning and design tasks involving intermittent contact. In this talk, I will present a basic underlying multibody model and discuss its use in solving a range of problems from planning dexterous manipulation to the design plate motions for vibratory manipulation. As will be seen, the power of simulation-based approaches are that they can yield solutions when human intuition completely fails and they can provide better solutions by considering possibilities outside of the designer’s “comfort zone.”