Controlling Gaming Agents via Reactive Programs

摘要

This paper introduces Icarus, a language for specifying the behavior of artificial agents that require extremely reactive, yet goal-focused control programs which include non-trivial cognitive elements. The key features of Icarus are the ability to express hierarchical objectives, actions, and plans, the use of function call semantics, and the ability to express processes as objectives, all embedded in a reactive control loop than considers every relevant action on every cycle of the interpreter. Icarus has a small footprint and executes fast. We use a body of examples to illustrate these features, and establish Icarus' relevance to the problem of encoding purposive, but reactive artificial agents for computer games.