GridWorld

1
GridWorld Introduction

• Peter Cowling University of Bradford
• P.I.Cowling_at_bradford.ac.uk

2
Outline

• Aims of the project
• Design Philosophy
• Outcomes
• The student experience
• Teaching/assessment methods
• The teacher experience
• Demo and Conclusions

3
GridWorld - Principal Aims

• Make learning fun and competitive
• In the following areas
• AI software design and implementation
• C.NET and VS.NET
• By getting students to write AI software agents
  for games
• and continuously running a competition between
  AI agents so they can see how each agent is
  doing.
• as for Virus and Terrarium

4
GridWorld Further Aims

• but also offer an open platform for easily
  creating new games
• and enable the development of a simple but
  flexible API for each game
• and make this robust
• using the idea of a grid environment which can
  be used to represent most board, card and video
  games.

5
Design Philosophy

• GridWorld Engine
• Client
• Form
• Graphics rendering
• File I/O
• Server (part of client)
• Results
• Security
• Web
• Game Manager

• Game
• Rules
• Images
• (Results Data)
• (User Input)

6
AI for GamesCM-0328D

• Prof Peter Cowling
• Sponsored by Microsoft
• Software co-developers Black Marble
• http//www.comp.brad.ac.uk/intranet/modules/AIFG/

Tuesday 2-4 Richmond N3 (from 9th Feb) Monday 2-4
Chesham C3.10 (from 2nd Feb) External speaker
Black Marble (Mon 2nd Feb 12-2 Horton D2.06)
7
From 2007
8

• Research
• "A Coevolutionary Model for the Virus Game", P.I.
  Cowling, M.H. Naveed, M.A. Hossain, IEEE
  Symposium on Computational Intelligence and Games
  (CIG 2006) (IEEE Computer Society press), 45-51.
• Board Evaluation for the Virus Game", P.I.
  Cowling, IEEE Symposium on Computational
  Intelligence and Games (CIG 2005) (IEEE Computer
  Society press), 59-65.
• and Knowledge Transfer
• Soon-to-appear article in Times Higher Education
  supplement
• "Writing AI as Sport", P. I. Cowling, AI Game
  Programming Wisdom 3, Charles River Media (2005)
  89-96.
• "ViewpointAI Thinking Ten Years Ahead", P.I.
  Cowling, invited article in Develop magazine
  (Nov/Dec 2005).
• "Using Bugs and Viruses to Teach Artificial
  Intelligence", Peter Cowling, Richard Fennell,
  Robert Hogg, Gavin King, Paul Rhodes, Nick
  Sephton, International Conference on Computer
  Games Artificial Intelligence, Design and
  Education (CGAIDE 2004), Microsoft Campus,
  Reading, UK, 8-10 Nov 2004).

9
The Student Experience

• All learning is hands-on.
• Students learn new ideas through using them.
• Student motivation is very high
• Many students comment that this is the most
  enjoyable module they have experienced
• Some students who are unsure of their programming
  skills produce excellent software
• Students recognise the career benefits of using
  up to date technology
• Students competitive drive is harnessed
• Student comments speak for themselves

10
Presentation of Material

• Few PowerPoint slides
• Code Talk
• Q A in lectures
• Detailed web pages with many links
• Weekly labs initially directed finally advice
  surgeries
• Prizes
• Focus (carefully) on the competition

11
AI subjects taught

• OXO
• AI design as thinking about thinking
• Simple rule-based approaches
• Scripting
• Board evaluation
• Minimax search
• Infectious!
• Positional features
• Pattern matching
• Alpha-beta pruning

12
AI subjects taught

• Infectious
• Parameter Tuning
• Genetic and Evolutionary Algorithms
• Testing and refinement for AI
• Bugs
• State machines
• Hierarchical State machines
• Multiple Concurrent State machines
• Avoiding deadlock
• Opponent modelling
• Pathfinding and A
• Terrain Analysis

13
Subjects considered/to be taught

• Games design
• Reinforcement Learning
• Neural Networks
• Agent coordination
• Game theory
• Bayesian Probability
• Data Mining
• Mapping

• Fun metrics
• Behaviour capture
• Emotion
• Environmental triggers
• Behaviour hierarchies
• Dynamic scripting

14
Use in Assessment

• Coursework 1 (40 of module marks)
• Design, build, test, refine and document a board
  evaluation function for Infectious!
• Coursework 2 (60 of module marks)
• Design, build, test, refine and document an agent
  for Bugs
• Marks awarded for originality, design depth,
  quality of testing, refinement based on test
  results, documentation

15
The Teacher Experience

• Good attendance in body and mind
• Lots of questions and interaction
• Coding in lectures quite nerve-wracking and very
  instructive to students
• Students frequently brought ideas from outside
  the course
• Enjoyable teaching experience

16
Conclusions

• Game AI is a good medium for teaching AI as well
  as programming/debugging and software design
• and encourages students to think outside the
  box.
• Hands-on with coursework-only assessment works.

17
Demo and Questions