Cognitive Architecture

AAMAS Tutorial (2016)

A half-day tutorial on Sigma was given at the International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2016) in Singapore on the morning of May 9.

The Sigma Cognitive Architecture and System

Sigma (Σ) is a nascent cognitive system – i.e., the beginnings of an integrated computational model of intelligent behavior – built around an eponymous cognitive architecture (a hypothesis about the fixed structure underlying cognition).   As such, it is intended ultimately to support the real-time needs of intelligent agents, robots and virtual humans.   Its development is driven by four desiderata – grand unification, generic cognition, functional elegance, and sufficient efficiency – plus a unique blend of ideas from over thirty years of independent work in the areas of cognitive architectures and graphical models.  Work to date on Sigma has covered aspects of learning and memory, perception and attention, reasoning and problem solving, speech and language, and social and affective cognition.  It has also involved the development of multiple distinct types of intelligent agents and virtual humans.  This tutorial covers the rationale behind Sigma, the basics of its design and operation, and a number of the results that have been generated to date with it.

Presenters:

Paul S. Rosenbloom is Professor of Computer Science at the University of Southern California and Director for Cognitive Architecture Research at USC’s Institute for Creative Technologies. He was a key member of USC’s Information Sciences Institute for two decades, leading new directions activities over the second decade, and finishing his time there as Deputy Director. Earlier he was on the faculty at Carnegie Mellon University (where he had also received his MS and PhD in computer science) and Stanford University (where he had also received his BS in mathematical sciences, with distinction). His research concentrates on cognitive architectures – models of the fixed structure underlying minds, whether natural or artificial – and on understanding the nature, structure and stature of computing as a scientific domain. He is: a Fellow of both the Association for the Advancement of Artificial Intelligence (AAAI) and the Cognitive Science Society; the co-developer of Soar, one of the longest standing and most well developed cognitive architectures, during much of its early evolution; the primary developer of Sigma, which blends insights from earlier architectures such as Soar with ideas from graphical models; and the author of On Computing: The Fourth Great Scientific Domain (MIT Press, 2012).

Volkan Ustun is an artificial intelligence researcher in the Cognitive Architecture group at ICT. His general research interests are computational cognitive models and simulation. He is currently involved in the various aspects of the development of the Sigma cognitive architecture. Volkan has B.S. and M.S. degrees from METU, Turkey and a Ph.D. degree from Auburn University in industrial and systems engineering. Papers out of his Ph.D. research are recipients of The Gene Newman Award of Excellence in Modeling and Simulation Research and the Winter Simulation Conference I-Sim/ACM-SIGSIM Best Computer Science Focused Student Paper Award. He has also been selected as the Most Outstanding Graduate International Student
at Auburn University.

Outline

The outline below is what was planned, although many of the “additional topics” needed to be trimmed to fit within the time available.  For the hands-on portion of the tutorial, instructions will be posted later concerning how to install Sigma and to access the code for the examples.

  • Introduction
    • Cognitive architecture
    • Sigma desiderata
    • Graphical architecture hypothesis
    • Graphical models
    • The structure of Sigma (+ cognitive cycle & tri-level control)
  • Hands on (Sigma as an agent on a grid: A sequence of random walks)
    • Operators (+ conditionals)
    • Operator selection
    • Internal action execution (+ types & predicates)
    • Trials
    • External action execution (+ perception & action)
    • Value selection
    • External objects
    • Learning (of Maps)
    • Simultaneous Localization and Mapping (SLAM)
    • Semantic Memory (& Learning)
    • SLAM + semantic memory
    • Action modeling (& templates)
    • Perception modeling
    • Reinforcement learning
  • Additional topics
    • Rule memory (& mapping to graphical models)
    • Mental imagery
    • Distributed vectors (word embeddings)
    • Episodic Memory
    • Appraisal and attention
    • Theory of Mind (& multiagent systems)
    • Interactive, adaptive virtual humans
  • Summary

Tutorial Notes

The actual tutorial slides can be found at AAMAS 2016 Tutorial Slides.  A page with instructions for the hands on portion of the tutorial can be found at AAMAS 2016 Hands On Tutorial Instructions.

Related Reading Material

Many papers on Sigma can be found at the Cognitive Architecture recent publications page (accompanied by brief blurbs on the various research areas). A good place to start is the 2013 AISB Quarterly short overview.  From there, the choice is very much a function of which topics are of particular interest.  There is a draft of a long article (~90 pages) that goes into quite a bit of detail on many of these topics, but it is not yet ready for public distribution.  A pointer will be added to it here when it becomes available.

Acknowledgement

This material is based upon research supported by the Army, the Office of Naval Research, and the Air Force Office of Scientific Research.