Summary

The situations and the progression of the game are taken into account by the strategies:

• Factors of the decision include enemy status, player state, and terrain layout.

• Features are equations capturing these factors, used to model the problem. Features should represent concepts consistently within unit range.

Decisions need to be made according to various objectives:

• There are usually global objectives, but players have the freedom to interpret them according to their mood.

• Moods guide the intelligent behaviors by providing a fitness function, or criteria to evaluate performance.

The result of the decision is a tactical behavior, appropriate for the situation:

• Tactical behaviors can be considered as compositions of components.

• The existing functional components (for instance, movement or shooting) can be customized for each tactical behavior.

Another AI component is needed to bring all these concepts together:

• The actual decision-making process uses features and criteria to pick the correct behavior. This can be achieved with experience or deduction, or arbitrary combinations thereof.

The next chapter focuses on using this knowledge to create an AI with deathmatch strategies determined by design. The following chapters elaborate on this prototype by adding learning capabilities.

Practical Demo

The animat called Ratty uses the moods system as guidance for strategies. Ratty does not take these into account, but just tells other players what it's trying to do. The debugging log also reveals what criteria it's trying to satisfy. The binary is available on the web site at http://AiGameDev.com/ with the source code.