Maze Navigation

There is no problem solving in nature, just adaptation. A frog, evolutionary biologists may argue, is not ``a solution to the problem of being a frog'', and it is not appropriate to say for example that it has solved the problem of jumping, or ``being green''. In the same vein, none of the behaviors we have observed among Tron agents have evolved deliberately, but rather as a consequence of the fight for survival.

Tron agents perform known basic robotics behaviors, such as obstacle avoidance or wall following, not because they have solved those problems, but rather as a consequence of evolving for survival.

Figure 3.42 shows a Tron agent that has been inserted on a maze. This agent visits some of the various passages, backtracks, then finally finds the exit and leaves. Approximately one every 10 evolved agents is able to find the exit to this particular maze. But the reason they do so is a need for survival, for there was never any payoff to do it.

The fact that the agent actually leaves the maze through the exit is particularly interesting. Why leave? Humans observing this behavior might interpret that there is a will to escape, a need for freedom; but on a deeper sense these are just consequences, as all animal behaviors are, of their being adapted to a competitive environment.

  

Figure 3.42: A Tron agent navigates out of a maze. It uses its own traced movements to explore until it finds the exit. There was never a purpose or explicit fitness reward for navigating mazes; instead, this agent behaves in this way as a result of its adaptation to the environment of Tron. The agent leaves the maze through the exit only because it has evolved to avoid confined spaces.