ABSTRACT
One of the central elements of any robust AI system is knowledge representation. This encompasses a wide variety of techniques and mechanisms for storing and accessing information about an AI agent’s perception of the world around it. The more powerful the knowledge representation, the more effective the AI can be. Propagation methods vary by application. For instance, a map used to represent the probability of an enemy occupying a given area might use a simple combination of setting high values when an enemy is spotted and allowing values to decay and “spread outward” to nearby points over time. Some applications need not perform any propagation at all, such as tactical maps that store the instantaneous locations of various agents. Such maps can be queried directly for as long as their information is deemed up to date, and when the map becomes stale, it can simply be wiped and recomputed from scratch.
