During simulated low-altitude flight, participants' control of speed is based on the optical flow rate ... forward speed over a planar surface, decreases in altitude.
Attentional Locus and Ground Dominance in Control of Speed During Low Altitude Flight Eric J. Adamic, Joseph Behre, & Brian P. Dyre University of Idaho Moscow, Idaho During simulated low-altitude flight, participants’ control of speed is based on the optical flow rate projected by the ground, even when changes in altitude make this information unreliable and other sources of speed information, such as the flow rate of a cloud layer above the flight path, could help to provide more valid information (Wotring, 2008). This study examined whether this ground bias in perceiving speed could be overcome by using a secondary visual search task to manipulate the attentional focus away from the ground. A task requiring participants to visually scan for potentially-colliding planes either above or below the horizon was coupled with a speed maintenance task similar to the task used by Bennett, Flach, McEwen & Russell (2006) and Wotring, Dyre, & Behre (2008). We found that altitude disturbances induced inappropriate speed control in a similar manner independent of whether the secondary task required the attentional focus to be directed above or below the horizon. These results suggest that ground bias in speed control is robust even when attention is directed above the horizon by a secondary visual task. INTRODUCTION Low-altitude fixed-wing flight presents a unique challenge to pilots due to a potentially deadly combination of a narrow margin for error in controlling speed and altitude and the potential for cross-talk between altitude changes and control of speed (Wotring, Dyre, & Behre, 2008). The leading explanation of altitude-speed cross-talk is that the visual system estimates speed primarily from global optical flow rate (GOFR): formally defined as forward speed divided by altitude or eye-height (Warren, 1982). For flight at constant forward speed over a planar surface, decreases in altitude result in increases in GOFR, while increases in altitude will result in decreases in GOFR. Hence, if GOFR is used to perceive and control forward speed, cross-talk between altitude changes and speed control will result. Using a speed maintenance task originally developed by Bennett, Flach, McEwen & Russell (2006), Wotring et al. (2008) observed exactly this type of speed-altitude cross-talk in simulations of low-altitude flight over a planar surface. This cross-talk is particularly dangerous for general aviation flight at low altitude because of the tradeoff between air speed and altitude dictated by gravity and aerodynamics. Because of the relatively low power of general aviation aircraft, altitude must be sacrificed to gain speed and visa-versa. Hence, it is critical to be aware of any changes in airspeed or altitude and make appropriate control inputs. Exacerbating this problem, flight displays providing unambiguous information about speed and altitude may be sampled less often at low altitude, because pilots are spending more time visually scanning outside the cockpit for obstacles. In principle, changes in altitude are optically decoupled from GOFR if a second textured planar surface (e.g., a bank of clouds) is located at an altitude equal to twice the altitude of the airplane. For such an environment, GOFR remains constant during altitude changes because as the plane gains altitude, the increasing flow rate of the clouds compensates for the decreasing flow rate of the ground. When the plane loses altitude, the opposite occurs. Thus, if speed is being perceived
and controlled based on GOFR, speed-altitude cross talk should be greatly reduced or eliminated by such an environment. However, Wotring (2008) found that in a visual flight simulation adding this second planar surface representing clouds at twice the plane’s altitude did not significantly reduce speed-altitude cross-talk. Participants appeared to exhibit a ground dominance effect and simply ignored the cloud layer whenever the ground was present, even though they would effectively use the cloud layer for controlling speed when the ground was not visible. Similar ground dominance effects have been observed in other visual tasks. Bian, Braunstein, & Andersen (2006) found that ground dominance affects the perception of relative distance to objects in a three-dimensional (3D) scene. Ozkan and Braunstein (2009) found that ground dominance affects binocular rivalry, where the ground surface presented to one eye suppresses a representation of a ceiling surface presented to the other eye. Seno & Sato (2008) found stronger vection (perception of self-motion) responses from movement relative to a ground surface as compared to a ceiling surface with the same optical characteristics. Interestingly, they found ground dominance for vection even when the observer was inverted so that the ground surface appeared in the upper visual field. The aim of the present study was to examine if ground dominance in perception and control of speed could be overcome by manipulating the locus of attention away from the ground to other elements in the visual scene. We used a visual flight simulation and speed-maintenance task similar to Wotring (2008) that included a simulated environment consisting of two planar surfaces located at equal distances above (representing clouds) and below (representing the ground) the simulated initial altitude of the participant. In addition, a secondary visual search task was included that required participants to visually scan the environment for other planes approaching from ahead either above or below the horizon and determine whether the plane was on a collision course or would pass by as a near-miss. Hence, the
(a)
(b)
Figure 1. Screen captures of actual displays showing the ground and cloud textures and the planes used in the collision detection task. Panel (a) represents an attention downward condition, and panel (b) represents an attention upward condition.
attentional locus was directed to either the ground plane below or the cloud plane above for the duration of each trial. If ground dominance found for perception and control of speed during low-altitude flight can be overcome by this attentional manipulation we predict two possible outcomes for when attention was directed upward toward the clouds: a) that ground dominance and upward attentional locus would cause both the ground and clouds to contribute to control of speed and therefore greatly reduce or eliminate the amplitude of cross-talk between altitude changes and speed control as compared to when attention was directed at the ground, or b) that the upward attentional locus would completely overcome ground dominance and participants would pay attention to the clouds only, resulting in no reduction in the amplitude of cross-talk, but instead a 180 degree phase shift (increases in altitude cause an increase in perceived speed) relative to when attention was directed to the ground (increases in altitude cause a decrease in perceived speed). If ground dominance is stronger than our attentional manipulation, however, we expect to find that speed-altitude cross-talk exists with the same amplitude and phase regardless of whether attention is directed upward or downward. METHOD Participants Ten university undergraduate psychology majors participated, representing a convenience sample of the psychology participant pool. All subjects were tested for 20/20 Snellen acuity and were naïve of the purpose and hypotheses of the experiment. One participant was excluded from the analysis because of a technical problem that occurred during experimental testing that compromised the data. To keep an equal number of participants per group, a second participant was chosen at random to be excluded from the other
experimental group, resulting in a total of eight participants’ data being analyzed. Stimuli and Apparatus The simulated environment consisted of planar surfaces 16, 000 m in width and 83,000 m in length representing the ground and clouds. Both surfaces were textured with a 256 x 256 pixel square consisting of a random swirling pattern, repeated 35 times and colored appropriately (green-black for ground and blue-white for clouds). At its edges, the pattern blended with adjoining squares to appear as a continuous, seamless texture (see Figure 1). The cloud surface was located at a constant altitude of 185.2 m, twice the initial altitude of the viewpoint. Similar to Wotring et al. (2008), sum-of-sines disturbances were defined by two interleaved sets of nonharmonically-related frequencies. The frequencies of these interleaved sets are listed in Table 2. The presence or absence (zero amplitude) of the altitude disturbance was manipulated within-subjects, while the assignment of the frequency sets to either the altitude or speed disturbance was manipulated between-subjects. As a result two groups (S
