 |
 |
 |
 |
| |
|||||
|
|
|
|
|
||||
Building Autonomous Cognitive Models of Air Traffic Controllers By Steven Estes, Chris Magrin, and Frank Sogandares Every year, hundreds of controllers and pilots participate in simulation exercises at MITRE's Integrated Air Traffic Management Laboratory. Each of them typically spends several hours to a week simulating their job tasks, such as maintaining aircraft separation. They participate in these simulations to evaluate the latest technologies and procedures. Model Behavior In the same way that a computer model of a bridge would allow the bridge designer to evaluate the soundness of his design before constructing it, a cognitive model allows a system designer to test the effect technologies will have on a human. For the bridge, the goal is to adapt the bridge to its environment. In the case of a cognitive model, we are trying to determine how the environment has to accommodate the human. A cognitive model does this by representing basic human capacities (e.g., motor movement, perception, and cognition) and constraining each of those in the same way humans are constrained. With a well-defined cognitive model, we can gain quantitative insights into how effective a proposed system is and how that system might be improved—without involving 10,000 volunteers.
Cognitive models embedded within a simulation environment allow for macro-level system results that are based on micro-level cognitive data of a type typically only gained in human-in-the-loop evaluations. This data allows us to evaluate the impact of a new technology on something much closer to a system-wide scale. Perhaps just as importantly, these types of evaluations can happen early in the design process (even before working prototypes of the concept to be tested are developed). Because the models identify potential problems in the design process, developers can make more effective changes and determine the viability of different concepts. The result is fewer design iterations and the ability to test with live controllers and pilots nearer to a concept's final form. A Model of Efficiency Beyond the goal of large, cognitive model-in-the-loop evaluations, cognitive models can be used within real-time environments to facilitate human-in-the-loop evaluations by acting as pilots, controllers, or traffic managers. They may also be used in smaller-scale human factors analyses of new technologies, determining the cognitive and temporal demands that the technology will place on the humans who use it. Cognitive models can run faster than real time, enabling data collection for thousands of controllers in less time than it would take for one human controller. The intended use of these models is not to replace the controllers and pilots who participate in MITRE's air traffic management simulations. There is no replacement for actual participants. But if we can use models—based on the thought processes from actual controllers and calibrated against actual controller behavior—early in the design process and in the context of large evaluations, cognitive models will be able to supplement their feedback with insights into how a new technology will affect not just a sector, but the entire national airspace system. The result: tools pilots and controllers can use to improve the safety and efficiency of the national airspace system. |
|
||||
For more information, please contact Steven Estes, Chris Magrin or Frank Sogandares using the employee directory. Page last updated: October 9, 2008 | Top of page |
|||||
Solutions That Make a Difference.® |
|
|
Download this issue [614KB]