A gestural instruction learning robot using information infrastructure

This paper proposes a gestural instruction learning algorithm for robots which move in response to video information. Applying the algorithm to an actual moving robot in a trajectory learning experiment confirms that it enables a robot to understand, on the same level that a dog might, both the meaning of a human macro sign (i.e. a figure-eight sign) and the qualitative sense inherent in a human macro qualitative instruction (i.e. a figure-eight trajectory with a large width). The proposed algorithm refines the robot moving trajectory through the use of a fuzzy associative memory system. It is demonstrated that the use of macro qualitative instructions in the proposed algorithm enables trajectory learning to be attained more quickly than with the use of micro instructions in a conventional algorithm