Daily Note: Agent Factory Framework

These notes are a summary of concepts presented in “Agent Factory : An Environment for Constructing and Visualising Agent Communities.”

O’Hare, G. M. P. (Greg M. P.), Rem Collier, J. Conlon, and S. Abbas. “Agent Factory : An Environment for Constructing and Visualising Agent Communities,” 1998.

  1. 1. Introduction to Agent Factory
    • Software environment for rapid design, delivery, and experimentation of Multi-Agent Systems
    • Enables investigation of agent community behavior
  2. Belief-Desire-Intention Architecture
    • Properties of agents represented using mental attitudes
      • Belief: Knowledge about the environment and self
      • Desire: Computational states the agent wants to maintain
      • Intention: Computational states the agent aims to achieve
  3. Core Components of Agent Factory
    • Distributed component allocation table
      • Details of all Agent Factory Environments and their components
      • Communication handled via
        • Local message-passing for intra-environment communication
        • TCP/IP sockets for inter-environment communication
    • Generic agent model
      • Agents are classified as social and intentional
      • Constituent components
        • Mental State: Belief-Commitment-Capability model with Commitment Rules
        • Communication Mechanism: Based on Speech Act Theory
        • Acquaintances Model: Tracks agent relationships
  4. Key Functional Components
    • Agent interpreter
      • Manages interplay of beliefs, commitments, and capabilities
    • Perception builder
      • Monitors environment to adopt/update beliefs
    • Resolution theorem prover
      • Queries and derives beliefs using SLD-Resolution
    • Commitment management system
      • Oversees commitment adoption/revision
    • Action generator
      • Executes operations for agent commitments
    • Scheduler
      • Controls agent execution sequence using a discrete linear time model
      • Supports custom scheduling through reusable hierarchical models
    • World model
      • Represents physical/virtual environments
      • Supports integration of non-Agent Factory systems
      • Includes a 3D world model for agent navigation
  5. Agent Factory Tools
    • Agent instantiation/viewer tool
      • Instantiates agents and monitors their dynamic behavior
      • Supports customization of agent views and mental states
    • Interface customization tool
      • Enables system developers to modify agent components and interfaces
    • Agent factory visualizer
      • Provides a 3D visualization of agents and their environment
      • Includes aerial and agent-specific views
      • Supports real-time collaborative actions and diverse agent representations
  6. Agent Design and Deployment
    • Agents are rational, social, and intentional
    • Positioned and oriented within a defined world
    • Activation/execution occurs after connecting agents to the world
    • Scheduler ensures commitments are honored in a time-grain framework
  7. Visualization and Real-Time Behavior
    • AFV displays avatars and collaborative agent behaviors
    • Users can select unique objects and agent behavior types
    • Provides flexibility in visualizing multi-agent systems via customizable views
  8. Key Functionalities of Agent Factory
    • Access to named nodes in the world model
    • Dynamic creation of VRML elements from string descriptions or URLs
    • Captures and initiates interactions between environment nodes