Tron (or light cycles) is a game of pursuit and evasion played on a computer screen. It appeared on the movie Tron in 1982.
Tron was chosen as a domain to investigate machine learning because it is a simple one yet presents several features that are known difficulties for machine learning: A dynamic, real-time environment where a perfect strategy is not known. Human and machine intelligence can show their particular weaknesses and strengths chasing each other in this competitive arena.
The purpose of this project is to be a testbase for doing experiments with state of the art computer learning paradigms such as evolutionary computation and artificial life. We used Tron as an introduction to the strengths, weaknesses and the open questions in this research area.
These are the rules of the game: Two opponents move at constant speed. They leave a trail behind, which is an impassable wall. The first player to crash against a wall loses. Only right-angle turns are allowed. The opposite edges of the screen communicate with each other.
A Tron game as seen on the computer screen:
Two players, blue and black, started near the
middle of the screen and chased each other.
The game finished when the black player
crashed (a red asterisk marks the site) while
blue was alive nearby