Abstract
Virtual worlds are inherently complex, dynamic and unpredictable in nature. The interface they provide to external agent systems consists of low-level events and primitive data. This introduces an information representation gap between virtual worlds and declarative BDI-based agent systems. As a result, BDI-based intelligent virtual agents (IVAs) are not capable of identifying the complex abstract situations unfolding in their surrounding environment. In this paper, we describe a two-step process that enables an IVA to identify the complex situations they encounter. First, complex event recognition mechanisms are applied on the low-level sensor data received by an IVA. Complex events identified in the first step are compared against a domain-specific situation model to identify active situations. The situation model helps the agent to be aware of the start and end of situations, and also to be aware of any active situation at any given time.