A theory for mentally developing robots

Abstract

This paper introduces a theory about mentally developing robots. The limitation of the traditional agent model is raised and a new SASE agent is proposed, based on our SAIL developmental robot. We formulate the manual development paradigm and autonomous development paradigm. The performance of a developmental robot is then formulated as reaching the norm of a human age group. The framework of autonomously generating brain representation is investigated in mathematical terms. Some techniques of such a representation are provided based on the author's SAIL-2 developmental algorithm (1996). We establish the conceptual limitation of symbolic representation and from the limitation we propose that no developmental robot can use a symbolic representation. Finally, the completeness of developmental robot is investigated conditioned on five factors.