| Situation-driven modeling is becoming one of the critical technologies of managing the behavior of dynamic multi-component systems in real or synthetic environments. The need for situation-driven modeling is motivated by the complexity of the managed systems as well the complexity of the events happening in the world, where the systems are situated. We can refer to several applications domains, where situation management has been or being proven a vital solution, including diagnostic situation analysis in telecommunication networks, battlespace situation control in military applications, and predictive situation analysis in dealing with natural or human-inflicted disasters.
Any situation management process predicates the development of a situational model, i.e. an understanding how an intelligent agent perceives the state of the entities of the world and the inter-entity relations in time. The situation model should be expressive enough in order to enable the agent to reason about the world situation and plan its actions, predict potential future threat situations, or as an opposite task, chain back to past situations.
There are several levels ob abstraction of situation modeling, including signal, data, and cognitive levels of situation modeling. The focus of this paper is on cognitive level of situation modeling. Our objective is to describe a general framework of situation modeling, identify the issues and challenges, review the existing methods, languages and tools of situation modeling, and note important tasks of future research.
First we define what is situation management and why situation management is an important task in different application areas. Then we introduce the basic concepts of situation modeling and the role of situation modeling in the overall task of situation management. A major portion of the presentation will be a critical review of main approaches to situation modeling, including logic-based situation calculus of states, actions and fluents; semantic network-based approaches such as frame-based and case-based situation modeling; and computational methods such as Bayesian network-based and game-theoretical approaches. We will also briefly review languages that have been used for situation programming like GOLOG, CLIPS, Prolog, OWL and DAML.
The final part of the presentation refers to open issues of situation modeling and outlines a future research agenda.
|