Graham Kendall
Individual Projects (3rd Year Undergraduates)
(Academic Year 1999/2000)
Introduction
This page details the third year individual projects that are/have been supervised
by Graham Kendall who is a lecturer in the Department
of Computer Science at The University of Nottingham.
This page contains details of projects supervised in the academic year 2001/02.
The main project page can be found here.
Downloadable Dissertations
Some of the projects are available for download, so that you can view them.
This facility is for current third year students, being supervised by me,
who can use previous dissertations for their own research and also to look at
previous dissertations to see how to lay them out etc. (it also saves the inevitable
requests I receive asking to see a "good" project from last year, with the risk
that I never see it again!).
As such, these files are password protected and, to get the username/password
you need to EMAIL me. People outside
of the university will not be given access.
In addition, I cannot give access to those students
outside of my tutor group. You should approach your own tutor to get copies
of previous dissertations they have supervised.
Note : I cannot make the files available in formats other than those
already supplied.
3rd Year Projects Supervised in 1999/00
- Evolving Collective Behaviour by Chris Ward
(PDF/Word)
Abstract : Collective behaviour refers to coordinated group motion,
common to many animals. The dynamics of a group can be seen as a distributed
model, each animal applying the same rule set. This study investigates the
use of an evolved controller to produce schooling behaviour. A set of artificial
creatures ‘live’ in an artificial world with hazards and food. Each creature
has an artificial neural network brain that controls these behaviours in different
situations. A genome encodes the network structure, which may be combined
using artificial evolution with another if a creature should choose to mate.
Prey and predators coevolve to produce sophisticated behaviour.
This project won the prize for the most outstanding project of the year
(across all of computer science, not just those I supervised)
An article arising from this project has also been published in Artificial
Life
(Ward C.R., Gobet F. & Kendall G. Evolving Collective Behavior in an Artificial
Ecology, Artificial Life 7(2), 191-209. 2001 (special issue on "Evolution
of Sensors in Nature, Hardware and Simulation"))
- Artificial Intelligence Chess with Adaptive Learning
by Glenn Whitwell (Word)
Abstract : This document is the culmination of one year’s work
as part of a final year assessed project for a Computer Science Single Honours
Degree. It contains a discussion of the artificial intelligence concept of
Adaptive Learning with Genetic Algorithms and of how they were
used in the developing or ‘evolving’ of a chess-playing computer program.
Through the implementation of game tree search techniques such as minimax,
alpha-beta pruning, evaluation quiescence, and genetic-based methods applied
to a population of evaluation functions, the chess program displayed an autonomous
improvement in performance after 4400 training games. This report details
areas such as the motivation for the project, the research involved in completing
the project, the design of the chess program, the implementation of the program,
the reasoning for the techniques used, a discussion of the results obtained,
the success of the project, and a discussion of the potential project extensions.
A paper arising from this project was accepted by CEC2001
(Kendall, G and Whitwell, G. An Evolutionary Approach for the Tuning of
a Chess Evaluation Function using Population Dynamics. In proceedings
of Congress on Evolutionary Computation 2001 (CEC'01), COEX Center, Seoul,
Korea, May 27-29, 2001, pp 995-1002 (an IEEE conference). (pdf:265KB)
- Implementation of the No Fit Polygon using Minkowski
Sums by Amar Qureshi
Abstract : The No Fit Polygon (NFP) is a commonly used geometric
operation in many industrial processes, such as two-dimensional irregular
stock cutting. There have been many approaches to calculating the NFP, which
have tackled convex shapes quite efficiently, but have failed to tackle non-convex
shapes. This project aims to implement an algorithm that calcuates the NFP
based on Minkowski sums. The resultant application of this project will calculate
the NFP for a set of user specified polygons, which can either be convex or
non-convex.
- Experiments in Simulated Soccer by Philip Smith
Abstract : Robotic Soccer (RoboCup) is the attempt by the Multiagent
System community to create teams of robotic agents (players) that can play
Association Football. This dissertation focuses exclusively on simulated Robotic
Soccer. The main focus of this work is the creation of a team of autonomous
agents each working to achieve a common goal; winning RoboCup games. A significant
sub-goal is to achieve good team structure and shape without having to explicity
code it. The decision-making framework of my agents is a system of layers
that control increasingly complex behaviours. The first layer controls the
agents individual footballing skills. The second coordinates these skills
in a way that is appropriate for the situation. The third layer controls the
second layer in a way that causes the emergence of team structure and shape.
The agents are implemented in C++ under Linux using a standard RoboCup agent
template. The final team was able to win games of RoboCup and had a flexible
formation structure that emerged as a result of the interactions between the
individual skills of each agent.
- Playing Poker Using an Adaptive Learning Algorithm by
Mark Willdig (Word)
Abstract : Using adaptive learning methods, specialised computer
programs can evolve into expert systems based on simple principles. Adaptive
learning involves creating a program that can adapt and evolve to certain
circumstances by mutating its internal structure. Previous work has shown
that by using adaptive learning, an agent can be created that plays poker
successfully, able to play in a number of different environments against a
variety of opponents.
This study uses the game of poker to show how agents can evolve and adapt,
and expands on previous work by introducing bluffing and deal rotation.
Internal parameters are manipualted as a direct result of how the evolving
player performs, values being altered according to numbers returned by a normal
distribution graph.
The results of this were that the agent learnt successfully, initially losing
through the training period before learning was complete, and then playing
to maximise winnings.
The system performed well against all opponents, and the learning curve of
the evolving player proves that the system can adapt successfully to the rules
of poker. Initial limitation of previous systems are overcome by successfully
introducing deal rotation and bluffing and two rounds of betting with a trade
of cards in between.
This shows how the evolutionary strategy of parameter adjustment by increasing
and decreasing internal learning values as the agent wins and loses, accomplishes
the original objectives and makes significant advances on previous work in
the area.
A paper arising from this project was accepted by AI'01
Kendall G. and Willdig M. . An Investigation of an Adaptive Poker player.
In proceedings of the 14th Australian Joint Conference on Artificial Intelligence
(AI'01), Adelaide, Australia, 10 - 14 December 2001, LNAI 2256, pp189-200,
Springer-Verlag, ISBN 2-540-42960-3 (pdf:52KB)
(word:227KB)
- Artificial Life by Detlef Brendle
Abstract : This project Artificial Life is based on the
former simulation of John Conways Game of Life, which is the most common
root in the vast number of A-Life simulations nowadays. The main differences
are that the project is concerned about two life forms on the same grid instead
of only one and the project can also evolve the species on the same living
space. The main purpose is to show how evolution effects the chance of survival
of an organism.
- Internet Financial Information Filtering Intelligent
Agent by Amanda Banks
Abstract : This dissertation documents the research and design
for a system using an intelligent agent that filetrs financial information
retrieved from the internet, given user specified keywords relating to a given
stock portfolio. The agent uses a learning algorithm to improve performance
of filtering results and takes feedback from the user based on these results.
The application is designed for the first time personal investors looking
to research a stock portfolio, and the primary specification therefore was
ease of use. The system uses an intelligent agent framework to implement the
application allowing for future expansion, and a Kohonen map algorithm is
used to implement the learning capability.
This document details research done into the area of intelligent agents, specifically
on their definition, considerations, current systems and design issues. A
full specification of the problem is given, and an overview of the design
for the application demonstrated, given the lack of standard methodology for
designing intelligent agents. Details of the implementation and how the application
works are discussed, and the problems encountered during this stage are also
examined. The results are produced and possible enhancements to the system
suggested. The dissertation is summarised with the lessons learnt from the
project.
EMAIL : gxk@cs.nott.ac.uk
Home page
Last Updated 15th June 2001