Agent-Based Framework for Discrete Entity Simulations

Abstract

Agent-based modeling represents a new way of simulating the interaction of objects with their environment and among themselves through communication. Agent technology incorporates many of the features of more traditional (Lagrangian and Eulerian) efforts but has become feasible in modeling more complex systems only recently. This paper describes the development of a general parallel agent-based modeling framework in C++ on Department of Defense High Performance Computing (HPC) machines at the U.S. Army Engineer Research and Development Center (ERDC) Information Technology Laboratory (ITL) in Vicksburg, MS. It provides background on the motivation behind agent-based modeling and how it extends traditional modeling techniques. Differences are identified, and the strengths and weaknesses of various modeling paradigms are explained. A short history of continuum and discrete model coupling is provided, followed by a description of how agent-based techniques can incorporate features of both Eulerian and Lagrangian models. The architecture of the ITL agent framework and the construction of the behavioral functions that excite or inhibit agent behavior are presented in detail. The hardware/software evolution path is described as the code goes from a small, single-threaded binary running on a Linux workstation utilizing database calls (to meet memory requirements) up to its successful translation as a parallel implementation on large HPC machines. Porting and scaling difficulties are fully explained. The framework is tested on an idealized ecological sandbox representing the Noyo River basin, California, and the virtual growth of submerged aquatic vegetation (SAVs) under hydraulic conditions driven by the ADH (ADaptive Hydrology) code output of the ERDC Coastal and Hydraulics Laboratory.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 01, 2006
Accession Number
ADA481937

Entities

People

  • Mark Cowan

Organizations

  • Engineer Research and Development Center

Tags

Communities of Interest

  • Energy and Power Technologies
  • Engineered Resilient Systems
  • Human Systems
  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Agent-Based Simulations
  • Algorithms
  • Complex Systems
  • Computers
  • Databases
  • Drainage Basins
  • Engineers
  • Environment
  • High Performance Computing
  • Mechanics
  • Operating Systems
  • Parallel Computing
  • Parallel Processing
  • Physics Laboratories
  • Plants
  • Research Facilities
  • Simulations

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Coastal Oceanography
  • Parallel and Distributed Computing.