All the techniques from this can be reused for previous problems or upcoming ones, and everyone is encouraged to see some suggested exercises on the web site at http://AiGameDev.com/.
This chapter provides a good set of techniques and methodologies to solve simple problems by using a custom representation. It also reveals how to use learning to create sequences of actions that are very easily edited manually.
However, as far as manually crafting behaviors is concerned, no technique comes close to the simplicity of finite-state machines. They provide a practically perfect representation for arbitrary sequences of actions, taking away this flaw from (fuzzy) rule-based systems. In the next part, we'll discuss ways to combine state machines with other techniques using heterogeneous hierarchies of components.
Our animats are also much more flexible now; in fact, they have an almost complete set of reactive behaviors. In the next part, we'll be making them a bit more realistic, giving them the kind of instincts that animals have. To do this, we'll attempt to model different aspects of emotions, which will lead to more believable behaviors.