Design and implementation of a system that generates assembly programs from visual recognition of human action sequences

The authors propose a teaching method, called 'teaching by showing' in which a human instructor shows an intended task to a robot by simply performing it himself. As an important point in realization of the method, a theory for visual recognition of human action sequences is presented. Then they give a full description of a system that observes an example task by vision and automatically generates a robot program to execute the task. Results of experiments on assembly tasks in blocks world are also presented.<>