Remote Control Towers: Detectability of Flight Maneuvers and Risk of False Decisions Derived from Dual Choice Decision Errors in a Videopanorama-based Controller Work Position

Kurzfassung

Future remote control of small low traffic airports (Remote Tower Operation, RTO) will rely on the replacement of the conventional control tower out-of-windows view by a panoramic digital reconstruction with high resolution and pan-tilt zoom video cameras as basic sensor system (e.g.[1][2][3]). State of the art technology for a digital videopanorama is presently limited by a visual resolution of typically 1/30° (2 arcmin) per pixel (UXGA and HD-format camera technology, 4 - 5 cameras with 45 - 60° vertical field-of-view and 180° - 200° horizontal panorama, focal width of 8 – 13 mm,) under good visibility conditions. This optimistic theoretical value which does not include contrast effects (i.e. modulation transfer function) is about half as good as the diffraction limited resolution of the human eye (1 arcmin = 1/60°). Here we show that even with the use of a manually controlled pan-tilt (analog) zoom camera (PTZ with PAL TV-resolution, selectable zoom factor setting Z = 2 - 16, viewing angle 26° - 3°) experiments under operational conditions indicate a significant increase of decision errors under RTO as compared to the out-of-windows view in the conventional control tower, and we quantify the corresponding discrimination difference by means of detection theory. As part of an operational passive shadow-mode test [4,5] eight air-traffic controllers observed various flight-maneuvers during airport circling (left Figure), like bank angle and altitude changes, and gear-up / gear-down situations. The response matrix of two-alternative decision tasks as obtained by measuring the hit (H) and false alarm (FA) rates when reporting, e.g. gear-up vs. gear-down situation during approach, were analysed using Bayes inference and detection theory with graphical presentation through ROC-curves. The latter method in principle allows for separating discriminability from response bias and was previously successfully used for deriving a video framerate requirement that minimizes prediction errors in a simulated landing scenario [6][7]. An example is shown in the right Figure. The present analysis provides an initial estimate of the increase of decision errors under RTO conditions quantified by a discriminability (d’-) decrease by a factor of up to 3 as a worst case, as compared to the conventional tower condition. A Bayes inference analysis based on the same measured H and FA conditional (a priori) probabilities provides a corresponding increase for risk of false decisions. This finding may be related to the factor 2 reduction of video camera pixel resolution (optimistic value, neglecting contrast) as compared to the visual resolution of the human eye. Theoretically, without regarding diffraction (iris effect!) and contrast limitations, with Z > 4 the PTZ should provide an improvement down to sub 1 arcmin (pixel) resolution, the limit of gear-down detectability at distance 2.5 km. In practice, however it appears not sufficient with manual PTZ control. However, the discriminability decrease under RTO work conditions might be reduced by selecting advanced HD-camera technology also for the PTZ together with usage of automatic zoom camera tracking via image processing or Modes-S data fusion, and by improved operator training.