On Optimization of Predictions in Ontology-Driven Situation Awareness

Abstract

Abstract. Systems supporting situation awareness in large-scale control systems, such as, e. g., encountered in the domain of road traffic management, pursue the vision of allowing human operators prevent critical situations. Recently, approaches have been proposed, which express situations, their constituting objects, and the relations in-between (e. g., road works causing a traffic jam), by means of domain-independent ontologies, allowing automatic prediction of future situations on basis of relation derivation. The resulting vast search space, however, could lead to unacceptable runtime performance and limited expressiveness of predictions. In this paper, we argue that both issues can be remedied by taking inherent characteristics of objects into account. For this, an ontology is proposed together with optimization rules, allowing to exploit such characteristics for optimizing predictions. A case study in the domain of road traffic management reveals that search space can be substantially reduced for many real-world situation evolutions, and thereby demonstrates the applicability of our approach.

Towards ontology-driven situation-aware disaster management

“…In UFO a situation represents a snapshot, an instantaneous "picture" of the structural aspects observed. Therefore, we agree with Guizzardi et al (2013) and Hoehndorf (2005) in which a situation is classified as an endurant, in contrast with Baumgartner and Retschitzegger (2006), which defines a situation as an occurrent. While situation is an endurant, situoid is a perdurant, since it has a temporal extension, its begin time point (when it is activated) is obtained from the end time point of the situation activation event.…”

“…Some of those systems rely on approaches based on the Endsley's Situation awareness theory (Endsley, 1995) for dynamic decision making in human factors field, which was especially adopted by the information fusion community (Foo & Ng, 2013). Furthermore, modelling of emergency situations gained attention of the formal ontology community with the main goal of reusing structured knowledge through core ontologies, such as in Baumgartner and Retschitzegger (2006) and Scherp et al (2011). Our work is related to those efforts, where conceptual foundations for situation play a prominent role in the development of SA applications.…”

“…Despite the numerous efforts to support the conceptualization of disaster situations through ontologies (Apisakmontri, Nantajeewarawat, Buranarach & Ikeda, 2014;Baumgartner & Retschitzegger, 2006;Casado et al, 2015;Fan & Zlatanova, 2011;Hristidis, Chen, Li, Luis & Deng, 2010;Kandel et al, 2014;Kayes, Han & Colman, 2015;Kokar, Matheus & Baclawski, 2009;Matheus, Kokar & Baclawski, 2003;Othman, Beydoun & Sugumaran, 2014;Stojanovic & Artikis, 2011;Xu, Nyerges & Nie, 2014), to the best of our knowledge, none of those efforts explored the challenges in ontological analysis that arise from this field. Examples of those challenges are: What are the basic information units (infons) in a disaster situation?…”

Making workflows situation aware

“…project and employ AllegroGraph 8 as a database backend that stores 7 www.prologics.at 8 Figure 6: Overview on the System's Architecture ontologies as well as ontology individuals and provides a performant infrastructure for the implementation of envisioned algorithms. For the realization of the web service interface, we plan to use Java with Jena 9 as a semantic middleware.…”

“…Furthermore, Smanchat et al [47] revealed that all but one of today's approaches to workflow adaptations only react on context changes, while only [38] support predictive workflow adaptation. ProFlow, adopts situation prediction techniques for qualitative spatio-temporal reasoning [8] for integrated situation and workflow prediction, allowing agents not only to access and browse predicted situations and workflow states, but also to adjust and vary predictions by incorporating domain knowledge, enabling what-if-analysis. Major emphasize is put on overcoming the limitations of related approaches from areas like qualitative neighborhood-based prediction [18], qualitative simulation [2] [12] and robot agent control [19], often needing manual modeling on a per domain, or even worse, per prediction basis.…”