next up previous
Next: Java Applet Up: Experimental Model Previous: System Architecture

  
Tron Agents

Tron agents perceive the world through sensors that evaluate the distance in pixels from the current position to the nearest obstacle in eight relative directions: front, back, left, right, front left, front right, back left and back right. Every sensor returns a maximum value of 1 for an immediate obstacle, a lower number for an obstacle further away, and 0 when there are no walls in sight (figs. 3.5 and 3.2).

  
Figure 3.5: A Tron agent perceives the environment through eight distance sensors.

\resizebox*{0.4\textwidth}{!}{\includegraphics{graph/sensors.eps}}


Each agent or ``robot'' is a small program, representing one Tron strategy, coded as a Genetic Programming (GP) s-expression[84], with terminals {A, B, \( \ldots \), H (the eight sensors) and \( \Re \) (random constants between 0 and 1)}, functions {+, -, * (arithmetic operations),% (safe division), IFLTE (if less or equal-then-else), RIGHT (turn right) and LEFT (turn left)}, maximum depth of 7 and maximum size of 512 tokens. An agent reads its sensors and evaluates its s-expression every third time step: if a RIGHT or LEFT function is output, the agent makes the corresponding turn; otherwise, it will keep going straight.

The simple sensory capabilities imply that an agent's view of the game is quite restricted: the position of the opponent is unknown, and so is the complete view of the game situation.

Tron agents have no state variables, so the behavior of the agent is purely reactive, based solely on 7 distance sensors3.1. Whereas humans may base their game decisions on topological considerations (e.g. ``is this region open or closed?''), or follow plans, the complexities of a robot's behavior must emerge from the interaction with the opponent along the game.


next up previous
Next: Java Applet Up: Experimental Model Previous: System Architecture
Pablo Funes
2001-05-08