A display device, being either a HMD or a computer display, when placed at a static distance from the observer will have an inherent display FOV. The geometric
field-of-view (GFOV) is defined as the virtual viewing volume as an input to the display device. In most VR applications a perspective projection is chosen such that depth cues are consistent with a users real-world view, so the GFOV will match the HMD FOV resulting in a one-to-one (unity) mapping. Altering FOV parameters to non-unity levels would result in distorting the scene: users are presented with a mini- or magnification of the actual image. An experiment revealed that subjects preferred a virtual scene with a GFOV that was 14.9% larger than the HMD FOV [71]. This optical distortion has an impact on the users’ scene perception such as distance estimation [72], which stems from the fact that a visual cue such as affordance (Section 2.1.2) has been affected.
Minification and FOV issues in aviation has been predominantly researched re- garding synthetic vision (display) systems (SVS) and remotely-operated vehicles where the operator only has as small display available [73–75]. For flight train- ing, this technique is not applicable because the optical distortions hinder transfer to the real world where negative training transfer as discussed in Section 3.1 is unacceptable.
2.4.1
Amplified Head Movements/Rotations
HMDs with head-tracking offer immersion to users by enabling them to look around the virtual scene by natural head rotations/movements. But the narrow FOV means more frequent and larger head movements are required to see parts of the environment where normal short eye movement in the real world would suffice [76]. As introduced in Section 1.1.2, desktop systems pose severe viewing angle constraints for head rotations unlike view slaved HMD systems. There are physical viewing angle limits for the user: turning ones head beyond these angles would mean the displayed image is no longer visible!
Amplifying physical head movements/rotations to the virtual camera effectively solves this by allowing head movements to view their virtual surroundings on a low FOV display instead of using a one-to-one mapping [27, 32]. Figure 2.11
illustrates the example of mapping a head movement to the right by an amount ‘d’ to a virtual camera movement of a gain factor ‘E times d’. Applying gains to all three head-axis rotations and three movement/translatations allow for a full six degrees-of-freedom. While rotating the head in the real world, sensory information (Section2.1) such as vestibular, proprioceptive as well as visual information create consistent sensory cues that indicate self-motion (Section 2.1.4). Amplification
deceives the user into seeing a (virtual) scene that has turned or moved further than the actual physical head, but by exploiting the dominance of the visual cue above the vestibular confirmatory cue and the adaptability of the vestibulo-ocular reflex (VOR) as explained in Section 2.1.3.
Figure 2.11: Example of amplified head movement [77]
This technique has been shown both to be acceptable and useful: users preferred an average amplification of 1.26 while using a HMD with 60◦ FOV [31]. Sub- jects also found amplification natural: they noticed head movement attenuations significantly faster than amplifications [32, 78]. Scene movement in the opposite direction of head rotations (i.e. amplifications) should be avoided [79].
The inherent benefit of controlling the viewpoint with one’s head is intuitive [80], as humans already use their head in daily real life to look around. Furthermore, off- loading this control from other traditional input devices such as a mouse, remote control, game-pad or hat switch enables the use of input devices for their originally intended function [32]. Looking around an aircraft whilst securely strapped in the cockpit means that head rotations rather than head-translational movements are of primary interest. As such, amplified head rotations have also been implemented in a variety of VR settings other than HMDs for visual search tasks: desktop systems [28], fishtank VR [29], and surround-screen displays [30].
Conversely, the majority of HMD systems had tracking errors and latency issues, resulting in severe vestibular and proprioceptive cue mismatch. Investigating the effect of latency on perceptual stability, researchers found that subjects were not able to detect small inconsistencies between real and virtual yaw rotations [31,79]. Results showed that users may judge the virtual world as stable when the virtual rotation is slightly amplified compared to the real yaw rotation [79]. Furthermore,
users tend to unwittingly compensate for small inconsistencies between vision and vestibular sensation [71]. Again, this is attributed to the adaptability of the human VOR (Section2.1.3).
A current knowledge gap in the literature exists, with few studies addressing this emerging capability of amplification, and in particular to flight simulation applica- tions [26, 81, 82]. With the many different visual display configurations available in terms of size and viewing angles, there has been scarce documentation on inte- gration and usability [27]. Potential user sickness symptoms due to this technique in an applied piloting task are also unknown, with the majority of control studies being of short duration and not covering any active control tasks. After effects have also not been studied, an example of a yet to be researched topic is being how visual scanning patterns learned with amplification might inadvertently yield adverse effects. This is of interest for transfer of training to the actual operational environment.
2.4.2
Head-Coupled Factors
In order to document the evaluation process of integrating amplified head rotations into a flight simulator, lessons learned from prior research in head-coupled systems with HMDs is useful to systematically address common system design factors. A series of experimental studies have investigated the effects of lags in head-coupled systems on tasks involving tracking virtual targets with the head, tracking and manipulating virtual targets with the hands, simulated vehicle control, and target search and recognition [83]. It was shown that the target capture-time was signifi- cantly increased by imposing an additional lag of 67 ms on a head-coupled system which has a basic system lag of 40 ms for static targets. In an experiment, subjects were required to place the aiming reticle inside a target circle as quickly and as accurately as possible, and to keep it inside the circle for at least 350 ms. In an- other experiment involving the tracking of a continuously moving target, tracking errors were significantly increased by imposing an additional display lag of only 40 ms on a 40 ms basic system lag. The correlation between the head motion and the target motion also decreased with increasing lag, particularly at higher motion frequencies.
Another study on character search and recognition task in the HMD yaw axis reported consistent increases in search time with greater exponential lags in excess of 40 ms [84]. The frequency of head movements is also a factor influencing
the performance of a manual control task with a head-slaved display. In HMD experiments with a fixed FOV, participants were forced to turn their heads to keep the task in view but as system delay was increased, subjects increasingly inhibited their head movements because of the de-coupling between head position and the position of the displayed image [85].
These results stress the importance of system latency in virtual reality systems with head movement. It is imperative to keep system latency below 40 ms for head-tracking applications as the task difficulty increases substantially and may force users to adopt different strategies to cope with this factor. Assessment of the likely effects of image lags in a particular head-coupled system therefore requires a knowledge of the pattern of head movements required by users.