Temporal Representation and Reasoning for Dynamic Environments

Abstract

Most of real AI applications developed under dynamic environments have to interact with the external world, deal with imprecision of data and make estimations about the possible data occurrence at different instants of time. A temporal model suitable for this type of domains must provide a representation framework able to capture external observations, update this information in the internal state application and deduce how these changes influence the application evolution. Reasoning processes for dynamic domains are generally quite complex due to the imprecision and variability of data. This usually leads to situations where the available time to update all the necessary information before processing the following change is not enough. When this occurs the internal model is not more consistent with the external world thus leading to dysfunctions in the system.

This paper presents a suitable temporal model for applications running under dynamic environments. The proposed framework allows to keep the world model consistent with the external world as well as the prediction of future consequences. All reasoning algorithms are designed as a search process between two time-points allowing to obtain approximate responses for a temporal query instead of optimal long time-consuming solutions.