Hierarchical and Heterogenous Reactive Agents for Real-Time Applications

We present a generic concept for autonomous agents with reactive behavior based on situation recognition in real-time environments. Our approach facilitates behavior development through specialization of existing behavior types or weighted multiple inheritance in order to create new types. Additionally, the system allows for the simultaneous generation of hierarchical and semi-individual group organizations using specification and recursive or modulo-based patterns. Our framework is designed to support the creation of large numbers of secondary characters with individual and group behavior in simulation environments such as game engines. The engine allows for the specification of a maximal time-per-run in order to guarantee a minimal and constant frame-rate. We demonstrate the usefulness of our approach by various examples with up to hundreds of individuals