Developmentally cognitive robot vision

A generic methodology for designing developmental cognitive robot vision systems was proposed to add development capability to cognitive vision systems. Four issues were discussed: (1) human-like architecture suitable for different visual purposes, (2) efficient memory organization, forming, and indexing mechanism proper for emergence of machine imagery and experience-based learning, (3) life-like development model for open-ended cognitive improvement and perception development, and (4) engineering methodology to treat the robot intelligence embodiment as a life-like process combining evolution, development and learning mechanisms. Their possible solutions were also provided: (1) a so-called dual-visual system and dual-memory architecture framework, (2) an "interaction procedure recording" theory plus a new ecological neural evolution approach to solve the interaction and binding problem between dual visual systems, (3) a development model based on a novel hybrid intelligent system theory embedding memory and communication as its inner components, and (4) a four-phase robot life-span development schema for behavior engineering.