Contents
1. Introduction / PAGE
2. Demo Video
3. Project Details
4. Results
5. Conclusions
6. References
7. Team

Summary

Crash and Burn is a competitive tower defence game:

• A player wins by sending units across their opponent's grid to the goal
• The player and the opponent each have a grid that they are defending by placing towers on it
• Towers create a longer path for and destroy units

Introduction / PAGE

The artificial intelligence aspect of this game can be split up into 3 agents:

Tower: A tower is used to block units' paths and make them find a different route to the goal. It can also shoot projectiles at a unit. If a unit's path is blocked by towers, it will proceed to attack it.

• Percepts: in range units, surrounding towers, attacking units
• Actions: select target, shoot
• Goals: let the minimum number of units past it, do not get destroyed
• Environment: grid of game containing tower/unit/empty tiles, path of units, amount of units

Unit: A unit tries to traverse the grid of towers in order to make it to the goal. They determine a path through the towers based on distance, probability of survival, control flow, etc. If no path is available, they proceed to create a path by attacking and destroying towers.

• Percepts: tower layout, other units' paths, tower strength
• Actions: choose direction, move toward goal, attack
• Goals: reach the goal by traversing the grid of towers, do not get destroyed
• Environment: grid of game containing tower/unit/empty tiles, addition of towers at any given time

Opponent: The opponent is the AI that the player plays the game against. It is responsible for managing its resources, used to create towers and units. It lays down towers to block the players units' path to the goal, while creating units to traverse the player's own tower grid.

• Percepts: player's tower layout, player's units, resources count
• Actions: create unit, place tower, upgrade tower
• Goals: prevent player's units from traversing its grid, get required number of units through player's grid before player
• Environment: player's tower layout, path of player's units, amount of player's units

Demo Video

The player will show in this paragraph

Project Details

• Units: A* search to determine most efficient path, with adjusted heuristic to determine a safe path
• Towers: shoot closest unit until out of range or dead
• Opponent: builds towers on a pre-built path, sends units in groups to ensure some get through, use heuristics to determine whether to send units or place towers with resources

Results

• Tower and Unit Agents work as envisioned
• Opponent Agent was difficult because of spatial reasoning and developing logical heuristics
  • At the beginning of the game, the Opponent can be difficult to defeat
  • As the game progresses, it gets easier
  • Once the player has seen how the Opponent operates (a couple games), they can almost surely utilize this to win
    • This is due to a lack of variability in the Opponent's game play

Conclusions

• Computing search paths for a dynamic real-time environment is slow!
• Determining a build path for towers that makes sense takes a lot of effort.
• Main improvements that can be made:
  • Efficiency
  • Opponent heuristics refined

References

irrKlang
Shadows, Reflections, Lighting, Textures. Easy with OpenGL!
Shockwave 3D @ Lighthouse 3D
OpenGL @ Lighthouse 3D
NeHe Productions

Team

Tyler Casella is a 3rd year computer scientist at Cal Poly. His favorite graphic framework to work with is XNA, although OpenGL has been growing on him. Tyler also enjoys traveling through time to help old people of the past cross the street.

Andres (who goes by Andy, or in some circles "The Mexican"), is a coughFifthcough-year Computer Science major at Cal Poly who can often be recognized by his large hair and his rugged good looks. But don't let his manly scars fool you. He really is a nice guy. His hobbies include watching movies no one likes and making jokes no one gets. As the fearless leader of this intrepid group, he urged his team toward victory and it was glorious! Surely this is a great feat!

Ken Hung is a Computer Science student at Cal Poly. He has done extensive work with the models, particles, and animations for Crash and Burn.

Jim Langloid iaculis purus dui, et tempor sapien. Morbi ut libero ligula, eu pellentesque sem. In quis adipiscing velit. Donec id libero vitae massa pulvinar pulvinar ut ornare elit.

James Pearson is a scripting language junkie utterly in love with Python and Ruby. Since there is currently no need for those in Crash and Burn, he works on the AI opponent. James is also the creator and maintainer of the website.

Josh Robinson is a 3rd year Computer Science major at Cal Poly. He is graduating in December 2010. He likes to describe himself in 3rd person because it makes him sound cool.

© 2010 the Crash and Burn team