Abstract
The event calculus (EC) has been used previously to model the behaviour of individual buses in a public transport system. Using this for usefully answering questions about traffic conditions means getting high performance from the backward chaining behaviour of a typical event calculus inference mechanism written in a logic programming language. In seeking to find periods of congestion in the historical data as well as to detect congested traffic as it happens, we have tightly coupled an event calculus implementation with a geospatial database system. We thus go beyond the forms of caching or windowing typically used for managing such spatial data within an EC system.
Our approach can answer historical and real time queries using the same query engine and EC predicates, allowing us to combine complex event processing and longer-term database operations without needing to degrade the precision of results through the introduction of time windows. We are exploring its use on a dataset containing temporal-spatial locations of buses operating over an urban setting that spans several years.