CONFLICT RESOLUTION: AN EXPERIMENT IN AUTOMATED AIR TRAFFIC CONTROL (31-Mar-1969)

          CONFLICT RESOLUTION:
AN EXPE.RIMENT I N AUTOMATED AIR TRAFFIC CONTROL by
Robert Lieberman

Carnegie-Mellon University
Pittsburgh, Pennsylvania
March, 1969

Introduction

    A major step in the automation of a.ir traffic control is the development of a dependable, automatic conflict resolver. A t first glance, this goal seems unreachable, since re-routing of conflicting flight paths is contingent on performance characteristics of the planes involved, detailed knowledge of terrain and weather by the controller, and the existence of restricted airspaces. And it is true that when compared with a controller, an automated system w i l l probably fare poorly in these respects, as has been demonstrated by numerous attempts at truly complex problem solving by computer. However, it is likely

that a properly designed automated system can process fifty flight paths nearly as effectively as two. Such performance is impossible for the human controller, whose excellence as a processor decreases rapidly between two and fifty flight paths. Hence, in at least one sense, the feasibility and desirability of a conflict resolution system increases with the number of planes in an airspace.

    The important point is that a computer program's effectiveness
lies not in its complex problem solving ability (which tends to be
small), but in its ability to "juggle" data efficiently at a much
higher level of airspace saturation than a human would dare. Apart
from its psychological consequences for the controller (and reduction
of the trainee shortage in light of diminished mental risk), the . -

conflict resolver, by increasing air traffic capacity, would postpone

, the invocation of restricted access policies due to saturation.

           Current limits on air traffic are due to the control network's inability to handle the traffic rather than physical limitations of the airspace. The latter obstacle is, we hope, very far off. The automated conflict resolver can remove the former, artificial
.._ bottleneck, thus greatly increasing the 1.ifespan of the present, relatively unrestricted flight scheduling system.

     It must be obvious that the conflict resolver w i l l not be able to resolve a l l air conflicts. In those parts of an airspace for which it
.fails, the resolver can at least suggest partial soltuions, leaving actual solution of only a subproblem t o the human controller. In this manner, he can concentrate effort on small, complex problems, while leaving the large, but relatively simple ones to the conflict resolver. Since each component is processing data for which it is best suited, resources are being well managed, with a resulting increase in efficiency and capacity for the entire air traffic control system.

Design of the Confl...