From Cognitive Modeling to Robotics: How Research on Human Cognition and Computational Cognitive Architectures can be Applied to Robotics Problems

Abstract

Cognitive psychology and Artificial Intelligence (AI) have long been intertwined in the study of problem solving, learning, and perception. The early pioneers of AI, Herbert Simon and Allen Newell, drew as inspiration chess masters and from their study developed computer programs to mimic the problem solving abilities identified in chess masters. The understanding of chess strategies relied heavily upon characterizing the problem space as a combination of symbolic inference and statistical pattern matching, which allowed for a quick understanding of the environment by computer systems. Recently, robotics has emerged as an AI domain, and the problem space has proven a difficult one due to the sub-symbolic nature of the knowledge. As robotics has emerged as a field in AI, cognitive architecture researchers have continued to refine their understanding of cognition in new ways that allow for the duplication of human problem solving with limited resources. The goal of this manuscript is to inform the AI world of the successes cognitive architectures have produced with the hope that this knowledge can be transferred to AI, and more specifically, robotics.