A simulation may have an arbitrary number of controller, event, and measurement (CEM) objects. They are added to a master list of CEM objects (CEMList), which serves as a means of accessing the reference to the object when needed. Each CEM object has access to the simulation’s VehicleList and CEMList. CEM objects can include behavioral models of human agents, as well as measurement and discrete-event models. For example, for this project we use rudimentary pilot and air traffic controller CEM models that serve as wrappers for the higher fidelity MIDAS cognitive models of these agents. Also, the clear-air turbulence sensor, surveillance radar, and communication channel agents are CEM-type models. And a CEM-type discrete-event model is used to generate aircraft randomly as they enter the simulation over time.
The environmental controller and database (ECAD) type of object establishes axis definitions and allows for the inclusion of static atmospheric and terrain models as needed. The model of clear-air turbulence used for this project is based on an ECAD-type model.
Any number of input/output (I/O) objects may be used to provide data read/write capabilities.
Other types of objects in RFS include networking agents (to connect processes running on different computers), the simulation timer, and the overall simulation controller.
A simulation of air traffic control typically needs to include many types of models -- discrete event and continuous time, for example -- integrated into a hybrid system model. Rather than force all models to fit into one of these types, the RFS architecture only requires each agent to meet minimal interface standards. Specifically, each agent must update its state upon command, report the time of its next update (or, in the case of some measurement and discrete-event agents, the next time at which an update might occur conditionally upon other events), and identify whether its own update requires any other agents to also update. All other dynamics of the components can remain internal to their models, without requiring intervention by the larger simulation architecture. This internalism prevents the need to place fundamental restrictions on the type of model allowed in the simulation.
In association with the current project, the RFS architecture was extended to incorporate the High-Level Architecture (HLA) protocol, using a networking object that can network together simulations running on distributed processors. This can link multiple instantiations of RFS, as well as allow an RFS application to interact with other processes that share an HLA federate structure. This networking is transparent to the rest of the simulation and therefore does not require special design of the agents for networked simulations.
Human Performance Modeling for Aviation Technology R&D
The development and deployment of advanced air traffic management systems involves the integration of human operators with automated decision support tools and supporting communications, navigation, and surveillance systems. The performance of these integrated systems will depend on the performance not only of the physical systems but also of the human operators, including both the flight crews and air traffic controllers. The capabilities of the humans involved will determine not only the operational performance in terms of such measures as throughput or delays, but also the level of safety that can be achieved. Therefore, appropriate modeling of human performance is central to effective research programs directed at the development and evaluation of such systems.
While the established approach to addressing these human performance issues relies on real-time, human-in-the-loop simulation, as discussed in the Phase I report for this project (Bobick, et al., 1999), the costs and time involved in performing such simulation experiments present significant limitations to the number of scenarios that can be studied, as well as the number of repetitions of each experiment that can be run. This effectively limits the value of real-time simulation for any realistic systemwide safety assessment. While fast-time simulation reduces the cost of examining many different scenarios, and permits many repetitions of each experiment, the value of the results of this analysis for safety assessment depends entirely on how well the simulation reflects the behavior of the humans in the system.
中國航空網 www.k6050.com
航空翻譯 www.aviation.cn
本文鏈接地址：DEVELOPMENT OF FAST-TIME SIMULATION TECHNIQUES(15)
