University of Notre Dame
Browse

The Cost of Communication: Efficient Coordination in Multi-agent Territory Exploration Tasks

Download (3.02 MB)
thesis
posted on 2006-04-17, 00:00 authored by Paul William Schermerhorn
Some previous studies of the adaptiveness of communication for behavior coordination have found that communication is beneficial and, therefore, will evolve. Others point out shortcomings in prior studies with positive results and find that when those shortcomings are addressed, communication no longer will evolve for coordination. It appears, however, that communication has evolved for coordination, which suggests that those studies with negative findings also are missing some key factor. None of these studies undertakes a systematic examination of important variables such as communication range, sensory range, and environmental conditions. To address this shortcoming, an extensive series of simulations is presented that explore the effect of such parameters on the utility of communication for coordinating agent behaviors in the multi-agent territory exploration (MATE(n)) task. Agents in the MATE(n) task are required to visit all checkpoints given in the environment in as little time as possible; n agents are required to be present at a checkpoint simultaneously for it to be counted ``visited.' A comparison of the absolute performance of communicating and non-communicating agents on the MATE(n) task (i.e., their performance without regard to cost) finds that there are configurations in which communication provides an absolute performance advantage. A subsequent analysis of the results establishes constraints on the cost of communication must be satisfied in order for it to provide a benefit in relative performance} (i.e., absolute performance scaled by agent cost), required for communication to evolve. Further analysis determines that these cost constraints are probably too strict, making it unlikely that communication evolved for coordination in MATE(n) tasks.

History

Date Modified

2017-06-02

Defense Date

2006-04-07

Research Director(s)

Matthias Scheutz

Committee Members

Sunny Boyd Peter Bauer Gregory Madey

Degree

  • Doctor of Philosophy

Degree Level

  • Doctoral Dissertation

Language

  • English

Alternate Identifier

etd-04172006-100927

Publisher

University of Notre Dame

Additional Groups

  • Computer Science and Engineering

Program Name

  • Computer Science and Engineering

Usage metrics

    Dissertations

    Categories

    No categories selected

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC