Introducing robotic `common sense´ in real time dexterous manipulation

Describes the implementation of a real-time controller for dexterous manipulation with a hierarchical intelligent control architecture. The architecture is based on a blackboard model that enables incremental, opportunistic, and dynamic decision making and control. The following features have been incorporated into the dexterous robot hand controller: (1) controlling robot fingers using robotic common sense reasoning, based on qualitative representations of control variables and parameters; and (2) opportunistically switching between coarse qualitative control that is based on a physical model and conventional fine quantitative control that is based on a mathematical model. The controller was implemented on a multirobot grasping facility, and preliminary experimental results are presented