A Method for Learning Tasks at some Levels of Abstraction from Human Tasks in a Virtual Environment.
Toshitake TATENO, Masanori Igoshi · TRANSACTIONS OF THE JAPAN SOCIETY OF MECHANICAL ENGINEERS Series C · 1997
We report a method for teaching tasks based on human tasks in a virtual environment and making a database which is used for cases similar to the recorded environment. In order to describe the work environment, a method for representing configurational relationships is proposed. The representation method uses the qualitative states between two primitives. Since the representation is based on primitives, it can be combined with a parts connectivity graph, which can be used to analyze a machine from the topological point of view and represent the structur of a machine hierarchically. The tasks are hierarchically recorded as production rules using these representation methods, and when the structure of the work environment at any level agrees with the data, a new action can be created. Some learning case studies show that the method can be used to create new actions if the orientations and shapes of work objects are different from those in the teaching environment.