ABSTRACT: In the last several years, a collection of systems have been designed around cognitive psychological principles in an attempt to reproduce the thought patterns of the human mind. These cognitive architectures have made progress in modeling human thought through production system architectures; however, they traditionally have little interaction with the outside world. Additionally, these systems place a heavy burden on the modeler to hand code large memory structures. In this paper, we explore the architecture and capabilities of the Symbolic and Sub-symbolic Robotic Intelligence Control System (SS-RICS) which was inspired by cognitive architectures capable of higher order cognition such as the Adaptive Control of Thought-Rational. SS-RICS integrates production systems, semantic networks, machine learning and sub-symbolic processing to perform real-time mobile robot control. It is our belief that this integration of cognitive psychological techniques and AI techniques will allow mobile robots to have higher order understanding and interaction with the dynamic world in which we live.
Keywords: Cognitive Architectures, Mobile Robots, Semantic Networks, Production Systems, Hybrid Systems
1. Introduction
In the last several years, a collection of systems have been designed around cognitive psychological principles in an attempt to reproduce the thought patterns of the human mind. Such systems include the Adaptive Control of Thought Rational (ACT-R) (Anderson and Lebiere, 1998) and Soar (Newell, 1990). The primary goal of these systems is to model human mental processes, human knowledge, the nature of that knowledge, and how the knowledge is utilized and acquired. Soar and ACT-R have been designed on similar grounds in an attempt to define a unified theory of cognition and merge a number of cognitive phenomena into a single architecture (Newell, 1990). The semantic network Cyc fits into a more general category of knowledge representation systems in that its primary goal is to acquire a vast knowledge base spanning human consensus knowledge to build “programs with common sense” (Lenat, 1990). SS- RICS falls somewhere in the middle of these systems, as it is an attempt to merge large knowledge acquisition systems with cognitive psychological principles while utilizing traditional artificial intelligence (AI) techniques.
The AI community has struggled with the question of
how to reproduce the thought patterns of the human mind for years, resulting in two distinct paradigms (Sun, 2000). The first camp takes a psychological approach utilizing symbolic representations to drive intelligent systems and the second focuses on a mathematical approach utilizing distributed representations constructed with structures such as neural or connected networks. This clash was summarized particularly well by Minsky in an MIT-AI Laboratory Memo written in the early seventies.
“Workers from psychology inherit stronger desires to minimize the variety of assumed mechanisms. I believe this leads to attempts to extract more performance from fewer "basic mechanisms" than is reasonable. Such theories especially neglect mechanisms of procedure control and explicit representations of processes. On the other side, workers in Artificial Intelligence have perhaps focused too sharply on just such questions. Neither have they given enough attention to the structure of knowledge, especially procedural knowledge.” (Minsky 1974)
DOWNLOAD...
*Eric Avery, Troy Kelley, *Dr. Darush Davani
*Department of Computer and Information Sciences Towson State University, Towson, MD 21204
eavery1@towson.edu, ddavani@ towson.edu
US Army Research Laboratory Human Research and Engineering Directorate Aberdeen Proving Ground,
Aberdeen, MD 21005 tkelley@arl.army.mil
