Applying temporal databases to HLA data collection and analysis
Proceedings of the 30th conference on Winter simulation
The Multi-Agent Data Collection in HLA-Based Simulation System
Proceedings of the 21st International Workshop on Principles of Advanced and Distributed Simulation
Situation recognition: representation and algorithms
IJCAI'93 Proceedings of the 13th international joint conference on Artifical intelligence - Volume 1
Hi-index | 0.00 |
Our goal is to add new observations and analysis components in existing HLA distributed simulations for new behavioural studies without modifying the existing components. To this end, we design a method to enhance the analysis results, for example with new performance measure or hazardous situation detections of an existing simulation and hence find new ways of exploitation for existing simulation. Let us note that the measures we introduce concern not only the simulation components, but also the cooperation between them (which is seldom considered to our knowledge). The analysis of components cooperation requires to perform on-line analysis. The observation and analysis components use activity recognition techniques. All behaviors may not be directly observable, particularly when they are internal to a component, but it is always possible to consider their achievements in the form of activities, which are observable by definition. We assume that the various exchanges between components of a given simulation characterise the global behaviour of the simulation. In the HLA context, this is the case, thanks to the normalisation of these exchanges. Moreover, they are observable, and may hence be used as an event source. Those activities, and particularly the logical and temporal relationships between their events are described using chronicles. Briefly, the analysis relies on chronicles using as events the exchanges between the various components of the simulation, which allows to identify the activities which are looked for. To realise those observation and analysis components, we use two tools developed at Onera: Genesis (a platform for designing and developing HLA applications) and CRS/Onera (Chronicle Recognition System: a tool for describing and recognising chronicles), and we combine their approaches by extending Genesis federates with a chronicle recognition library built with CRS/Onera. To illustrate our tool relevance, we build an analysis component for an existing airport ground traffic HLA simulation: Future Airport. Future Airport is an Onera federative R&D project to provide an open and flexible modelling and simulation infrastructure to evaluate future airport operational concepts. To extend the existing results of this simulation we are adding two new measures. In this example, these measures evaluate a new performance metric of the traffic module and the planning module. First, we want to identify situations where planes spend more than 30% more than the average time in taxi-in or taxi-out. Second, we want to evaluate the minimum time between the detection of a hazardous situation and its occurence. Let us note that this analysis component gives new performance measures about the planning component and the traffic component without any modification of their source codes. We show that our method is not specific to this example, and can be extended to reuse and analyse any distributed HLA simulation and to any cooperation analysis on components.