A Representation Of Parallel Activity Based On Events, Structure, and Causality

Abstract

Most AI domain representations have been based on state-oriented world models. In this paper we present an event-based model that focuses on domain events (both atomic and nonatomic) and on the causal and temporal relationships among them. Emphasis is also placed on representing locations of activity and using them to structure the domain representation. Our model is based on first-order temporal logic, which has a well-understood semantics and has been employed extensively in concurrency theory. We show how temporal-logic constraints on event histories (records of past activity) can facilitate the description of many of the complex synchronization properties of parallel, multiagent domains.