Building artificial intelligence for games is an extremely involved process, and the cost of developing AI for minor characters may be greater than the payoff in increased immersion. Affordances, used to create smart objects, reduce the complexity of character controllers. While affordances improve the complexity of individual controllers, the technique does not take advantage of environmental information to reduce the behavioral decision space. We present contextual affordances, which extend basic affordances to use context from object, agent, and environmental state to present only the most relevant actions to characters. We show how contextual affordances improve even random-decision agents, and discuss the complexity of building contextual objects.
展开▼
