As wildfires threaten Northwest America it becomes increasingly important to design ways to contain a fire soon after its ignition. Given a fire's origin, the environment, and a constrained amount of fire prevention material, we propose a sandbox modeling tool designed to minimize the damage of a wildfire. Fire propagates according to biased stochastic probability function on a 2 dimensional lattice in this agent based model. As the fire spreads, rational artificial agents are tasked with building fire breaks in real time. Given only spatial context, agents are allowed to act on sites in their immediate Von-Neumann neighborhood. By introducing different tune-able parameters,such as topography, fire strength and resource scarcity, users to compare different containment strategies to minimize cost and maximize post fire yields. Our hope is that simulations in this framework may inform real fire fighting efforts after a fire has begun. Ideas are inspired from Highly Optimized Tolerance models and other probabilistic cellular automata models.
*This is part of a final class project for CSYS302 and is a work in progress. It is a sandbox model whereby different CA strategies may be written and deployed. Please see our final report for further details. Current work includes increasing computation efficiency of simulations and model validation with a number of different real world spacial temporal fire propogation data sets.
Recorded simulations of the wildfire can be found at :