Student Perspective of GridWorld

1
Student Perspective of GridWorld
2
Introduction

• Roderick Baker, BSc (Hons). Computer Science.
• Thomas Randall, BSc (Hons). Cybernetics Virtual
  Systems.
• Sean Wallace, BSc (Hons). Interactive Systems
  Video Games Design.

3
Synopsis

• Development of AI agents for GridWorld.
• The virtues of GridWorld.
• Problems overcome in creating agents.
• GridWorld as a learning tool.
• Conclusion.
• Questions.

4
Visual Studio.NET

• Previous Experience.
• Usefulness for Creating AI.
• Simplifies Debugging.
• Ease of use with GridWorld.
• Allows the user to browse the API with ease and
  get the most out of it.

5
C The .NET Platform

• Simple and intuitive language for all levels of
  developer.
• Makes code simpler read and create.
• Completely Object Orientated.
• Has garbage collection.
• .NET offered a lot of libraries easing the
  creation of game AI.

6
GridWorld

• Easy interface for creating game AI.
• Allowed full control of the player(s) within game
  rules.
• Offered functions to ease development.
• Instant access to seeing the agent playing a
  game.
• Replay files allowed students to view their
  agents progress and analyse errors.

7
GridWorld

• Trace window allowed students to potentially view
  behaviour as it happened.
• The client offered a local server so students
  could compare agents.
• Online server results, allows students to compete
  and allows agents to gauge tactics.

8
Problems

• At the time GridWorld was still in development so
  some errors hindered creation.
• We were effectively testers for the software.
• Logical errors on our part.
• Keeping CPU time below the allowed limit.

9
GridWorld vs. Conventional Teaching Methods

• GridWorld was fun
• To use, develop for, and to play.
• Much more motivational and involving than other
  methods.
• Almost instant reward for hard work.
• GridWorld felt tailored for AI creation.
• It allows developers to simplify difficult AI
  concepts.

10
Conclusion

• Introduction.
• Development of AI agents for GridWorld.
• The virtues of GridWorld.
• Problems overcome in creating agents.
• GridWorld as a learning tool.

11
Any Questions?