The behavior described in the previous paragraphs can account for the immediate needs of the firefighters. Firefighters require some level of intelligence in their actions, including abilities to read the landscape, anticipate future behavior of the fire, pick optimal paths to firefighting locations, coordinate activities with other firefighters, and evaluate the success of their strategy on the fly. Techniques for simulating these kinds of decisions is a subject beyond the scope of this chapter. There is an extensive body of literature in adding intelligence to agents; for a discussion of some of the practical considerations of this subject see Deadman and Gimblett (1994); Gimblett et al. (1996a,b).
What has been presented here is an appropriate set of inputs that such an ``intelligent'' agent would require if turned loose on the cellsurface. However, even with a reasonable artificial intelligence engine attached to the firefighters, the simulation described will take on an unpredictable quality, where the outcome will be unable to be determined without actually running the simulation. This unpredictable aspect, even in the face of predictable behavior by any one agent, provides for the most exciting components of such a simulation.