The successful demonstration of a dynamic linkage between the two model systems, MIDAS and RFS, is a significant research accomplishment. MIDAS and RFS, built for different purposes and using different simulation technologies and computer systems, have been made to work together as a single system model for purposes of a practical analysis application. We have shown that explicit models of human performance can be integrated into a systemwide air traffic simulation and that their presence makes a difference to the results of the simulation.
HLA with its Run-Time Interface (RTI) appears to be a very efficient and flexible networking solution.
Some of the challenges encountered included, for example, networking three software systems (the RFS and MIDAS simulations and the RTI middleware) running on two hardware platforms, passing messages among the three that are understandable to each, and synchronizing the timing between two independently running simulations.
Agent-based simulation is a very effective tool for modeling large systems, including both Monte Carlo and deterministic simulations. In RFS, once all agents are defined and coded, the model becomes very easy to use, and its versatility and flexibility for modeling systems at various levels of focus and fidelity (breadth and depth) is only limited by the available agent models.
The learning curve to develop agents in RFS was found to be quite steep, particularly to customize previously developed agents to the needs of this scenario of multiple aircraft flying in and out of three nested turbulence areas with air-ground communications.
MIDAS has been developed to simulate the cognitive, sensory, and motor performance of human operators to a great degree of fidelity, precision, and accuracy in both fast-time and real-time (human-in-the-loop) experiments. Its design was tailored to applications in well-defined, narrowly focused research experiments, e.g., to assess the impacts on human operator performance of a new display layout or mechanical control device. For this reason, the MIDAS modeling of the chosen scenario's controller and flight crew tasks and procedures and developing the HLA federates to implement the link with RFS proved to be much more challenging than anticipated. None of the many technical and conceptual issues that arose was intractable, however -- only time consuming. Priority was given to effecting the dynamic linkage, so that the flight crew and controller modeling was only partially completed, although sufficiently so to demonstrate the linkage.
Fundamental design limitations in MIDAS limit the scalability and practicality of MIDAS as a tool to model human performance in systemwide analyses. For example, human operators cannot be created dynamically (i.e., during the simulation either conditionally based on the state of the system or at a later time than the start of the simulation) in that all operators must be initialized at the start of the simulation. Also, all operator tasks, procedures, or resource loadings (and any changes to them for alternative scenario assumptions) must be coded into the computer program rather than input as data. Further model development to relax these limitations is both technically feasible and needed.
As important a beginning as this is, it is only a beginning, akin to that first telephone call of Alexander Graham Bell down the hall to his assistant Watson. Much research and development remain to be done to achieve a system that is credible and useful in actual applications to real problems. Some recommended steps in that direction follow:
An essential step is to rigorously verify that the operation of the linked RFS-MIDAS system is internally consistent and that its behavior conforms to the agent specifications developed in the course of the research. It is important to note that some of the RFS and MIDAS developments recommended below would need to precede verification tests. 中國(guó)航空網(wǎng) www.k6050.com 航空翻譯 www.aviation.cn 本文鏈接地址：DEVELOPMENT OF FAST-TIME SIMULATION TECHNIQUES(7)
