The Task Support System
|
The Task Support System |
|
Synopsis: This page describes the Task Support System, part of an experimental avionics system designed to support military pilots.
Keywords: Task Support System, distributed artificial intelligence
Last update: 23 Jun 97
We became aware of the importance of the management of cockpit (or flightdeck) activities as we developed the Task Support System (TSS), part of an experimental avionics package for a military aircraft (Funk and Lind, 1991).
The purpose of the TSS was to help military pilots execute tasks quickly and correctly. But as we worked on the TSS, we found that it was just as important to help the pilot manage tasks, for even a well-performed task does little towards mission success if it is the wrong task or if higher priority tasks are neglected.
We developed the TSS for a simulator representing a single-seat military aircraft and its tactical environment. The simulator was implemented on a Silicon Graphics Iris computer.
Due to the complexity of the cockpit environment, we used methods of distributed artificial intelligence to implement the TSS. Its major components were intelligent agents: software modules that represented significant elements of the cockpit and its environment, having adequate declarative and procedural knowledge to deal with subsets of the problem domain. System agents represented aircraft subsystems. They monitored subsystem data and maintained declarative knowledge about the aircraft and its environment for other parts of the TSS. Task agents represented cockpit tasks. Each task agent had a knowledge base that helped it determine when a task should be performed and how to work cooperatively with the pilot to complete it successfully. High level task agents used their knowledge bases to help prioritize tasks. We implemented the TSS on an 80386-based personal computer in Objective-C, an object-oriented superset of the C programming language.
We evaluated the TSS in a simulator experiment in which 16 military pilots flew simulated missions in both baseline (no TSS) and enhanced (with TSS) cockpits. With the enhanced cockpit, overall task performance improved 38%, workload (as measured by NASA-TLX) was reduced by 13%, and pilot-perceived effectiveness improved by 83%. 81% of the pilots preferred the enhanced cockpit to the baseline.
Although most improvements were statistically significant, conclusive evidence of the success in improving activity managemente performance, especially in an operational setting, was not demonstrated. First, at that time, we had established no objective measures of CTM performance. Second, as the TSS was not the only element in the enhanced cockpit, it was not possible to separate out its effects. Third, the simulator environment was of low fidelity and the possibility of successful integration of the TSS into an operational aircraft was by no means assured.
Development of the Task Support System was supported by the US Naval Air Warfare Center - Weapons Division. Dale Robison, our technical monitor, not only provided valuable design guidance but also created the simulator and avionics environments necessary for the TSS.
Funk, K.H. and Lind, J.H., "Agent-Based Pilot-Vehicle Interfaces," IEEE Transactions. on Systems, Man, and Cybernetics, Vol. 22, No. 6, Nov/Dec 1992, pp. 1309-1322.