Video results: Using Neural Networks and the Genetic Algorithm for Multi-Agent System Control
Loading...
Uploaded by tjosgood on Apr 30, 2009
This project aims to show how genetic algorithms can be used to create a neural network capable of controlling a group of robotic agents for use in general problem solving. This was done by developing a complete simulation of a multi-agent system featuring a genetic algorithm and neural network subroutine.
The research builds on the work in a previous paper (Lee, 2003) about an experiment which used neural networks and genetic algorithms to control the behaviour of a multi-agent system allowing the behaviour of the agents to adapt and learn. The objective of the previous experiment is to co-ordinate a group of agents to track and capture a single target, the paper presents interesting ideas but lacks in any useful detail about how the results were obtained. The software developed for this paper aims to create a more general learning system that has independent simulation and genetic algorithm sections. This way it can be shown that it is possible to produce a neural network solution to solve any multi agent control problem.
The simulation software was written in stages, starting with the simple task of producing a neural network capable of replicating the behaviour of an XOR logic gate. Using this initial research the parameters for the genetic algorithm were investigated and tested to find a combination that offers good performance and consistency. This achievement alone demonstrates that the approach is flexible enough to produce to solutions to a wide range of computational problems.
After the genetic algorithm software had been tested and approved the XOR problem was replaced with the multi-agent system simulator in which the agents objective is to capture as many prey as possible by evolving new and more sophisticated tactics.
The end results were positive; the genetic algorithm successfully produced a control method capable of herding and capturing all 8 of the existing prey at the start of each simulation. The observed behaviour produced from random mutation and selection closely mimics that of natural predators and shows potential for great improvement and any number of other applications.
3:44
Virtual Creature Evolution Sequenceby kjlg7454,999 views- 3:21
Neural Networks Lesson 2: Probabilisitic Neural Networksby xoaxdotnet18,428 views
- 1:21
Genetic Algorithms TSPby ravivyatom17,888 views
- 10:01
Introduction to Neural Networks for C#(Class 3/16, Part 1/5) - hopfieldby HeatonResearch8,301 views
- 0:36
Dr. Mario, played by a 6th-generation genetic algorithmby captderiv15,881 views
- 7:02
genetic algorithms part 1by MantzosProtopapas8,910 views
- 9:46
Artificial Intelligence: This Time It's For Realby extensionmedia10,149 views
- 0:35
Evolving Truss Structures Using Genetic Algorithmsby ZongyiYang13,888 views
- 6:46
Genetic algorithms and TSP (2/2)by CarlSvenbro5,741 views
- 2:14
Genetic Algorithms: Evolving a human faceby ldbrandy29,609 views
- 8:56
Genetic algorithms and TSP (1/2)by CarlSvenbro5,336 views
- 1:13
Neural Network learning (recorded Java Applet)by mfunkmann25,001 views
- 9:52
Evolving Artificial Neural Networkby Etaanaru36,032 views
- 1:55
Neural Networks Classification with Sharky Neural Network (SNN) - shape "c|n"by SharkTimeCom4,101 views
- 3:53
Evolution of an Artificial Neural Network to control a crawling robot.by shaggyman8710,631 views
- 0:54
Genetic Algorithm (First test)by Spunkmeyer325,854 views
- 1:31
Feedforward Neural Network Training Using Backpropagationby fifothekid2,340 views
- 1:21
Multi Agent System Concept Demoby juliantkp1,560 views
- 1:01
Binary Decision Diagram (BDD)by onusbaum2,226 views
- 9:43
UAV Search and Destroy using Genetic Algorithmby nwharbour5,414 views
3:54 is the best one
echo0park 2 years ago