Date of Award

2019

Document Type

Thesis

Degree Name

Bachelors

Department

Natural Sciences

First Advisor

Doucette, John

Area of Concentration

Computer Science

Abstract

Communication is one variable of many in the implementation of artificial neural networks. This thesis explores the effects of adding communication to multiagent systems where each agent is an artificial neural network. We compare the performance of multiagent systems with full communication to the performance of multiagent systems with no communication to the performance of multiagent systems with one channel of communication. We measure the performance of each system using a simulation based off of real world objects where each of three small robots is an agent. The goal of these agents is to push a box over to a predefined area. The artificial neural networks controlling these robots evolve using an evolutionary algorithm - NeuroEvolution of Augmenting Topologies (NEAT). We collect fitness measurements over many generations and several trials then use multiple methods of statistical analysis to analyze the data.

Share

COinS