Kevin D. Reilly
Alan Sprague
Chrystian L Plachco

Agents and Model Abstractions in Intelligent Systems Simulations

Proc. Summer Simulation Conference, 2001



Abstract

The primary topic of this paper is agents and their potential in simulation. We approach the elusive agent with a postulation of working with definitions of three types, two at empirical and theoretical poles, and others in between. Examples and case studies are present to demonstrate contributions to model understanding and abstraction.


Addendum:

Short case studies illustrating agents and abstractions include: 1) White and Black Boxes 2) A NN (Neural Net) Model vs a LP (Logic Programming) Model 3)-4) Low-High Level Compartments I-II 5) Model and Its Animation Longer case studies include 1) External Memory Strategies, with agents as Localist NNs, as Distributed NNs, using Fuzzy Logic, using Logic Programming, using stochastic schemes, as response evaluators (objective and subjective); 2) a load balancing task, over N nodes, K monitoring agents with (limited) observation schedules but which share information gathering, animations of current and task performances.


Key Words: agents, model abstraction, neural networks, queueing