Some ideas about future air traffic control systems — theoretical and experimental investigations in a simulated environment

Abstract

It seems evident that a gradual and systematic computerization of the major tasks of Air Traffic Control (ATC) must be investigated and, if found feasible, introduced in routine operations. We have explored an area of Distributed Artificial Intelligence, called Distributed Planning and Problem Solving, as a reliable and effective methodology for future ATC. We discuss in this paper some basic concepts and terminology, and describe the Coordinator-Coworker Control Mechanism that can dynamically reorganize the planning system in response to the task at hand. The information structure supporting the control mechanism is a set of hierarchically organized Distributed Scratch Pads. This also provides the basis for self-repair in case individual nodes fail to operate, and allows for a graceful degradation of the whole system. We analyze the problems of Air Traffic Control and describe the kernel design of the airborne processors and explain the simulation-based planning process as well as its timing and selection considerations. The vehicle of our empirical studies, the Distributed Air Traffic Control Testbed is then introduced. We compare in it performance measures of three organizational structures: the Local Centralized Architecture, and the Location-Centered Cooperative Planning System with one- and two-level Coordinator-Coworker hierarchies. Finally, we present the results of a series of experiments performed in the Distributed Air Traffic Control Testbed.