5.5 Design Synthesis
5.5.4 Head-tracking Integration
With the main simulator specifications finalized, the remaining task is to integrate head-tracking to facilitate the view control augmentation (Section 2.4.1) used in the experiments described in Chapter4. Specifically, motion tracking systems most often derive pose estimates from electrical measurements of mechanical inertial, acoustic, magnetic, optical, and radio frequency sensors [176]. Because there is no universal single technology solution for all applications, each of their advantages and limitations need to be evaluated before adopting a particular solution [177]. For flight simulation, optical and inertial technology is most often used [178]. This is due to magnetic systems having too much latency and distortion from the metallic structures in the environment. Acoustic and radio trackers also require extensive calibration of the environment and are prone to external noise which is not practical for deployable solutions. With the low cost, commercial-off-the- shelf requirement in mind, an optical solution was selected and three candidate solutions found as shown in Figure 5.12.
The Kinect system [179] uses a combined 640×480 pixel camera system in the infrared and RGB color spectrum providing a 30 Hz update rate. Regular camera solutions like FaceTrackNoIR [180] utilizes face-tracking technology [181] to work with a wide variety of webcamera’s built-in laptops to stand-alone high-definition external cameras. The drawback with webcamera’s is that they require adequate
Figure 5.12: Optical tracking candidates (infrared, combination infrared/color camera and webcam based solutions)
lighting to function, which conflicts with the darkened environment requirement (1.R AppendixL) and during test struggled to maintain a steady 30 Hz update rate at the same resolution as the Kinect. The remaining and chosen solution was the TrackIR 5 [134] system, which delivered 120 Hz using passive infrared camera and position markers, shown in Figure 5.13. With a latency of just 9 ms, this was well within the 40 ms desired limits as found in Section 2.4.2. Furthermore, TrackIR had an established development track record with flight simulation software titles and provided the quickest integration route in this thesis. Prototype integration on actual military trainers as shown in Figure5.14 was further proof of the maturity and acceptance of this solution for flight simulators.
Figure 5.13: TrackIR system with infrared emitter and ballcap marker [182]
Since the first TrackIR system retailed in 2001, it has been used by consumers and researchers on a wide variety of configurations. Nevertheless, there was no documentation regarding fine-tuning the amplification parameters. The manufac- turer did supply default generic templates which formed the basis for in-house
Figure 5.14: Head-tracking demo on a part task trainer [183]
testing with the simulator. This required a learning curve and largely through trial-and-error feedback runs led to the used profiles as described in Section 4.2.3
and Section 4.3.3. Although the simulation software X-Plane, had built-in sup- port for TrackIR, a custom view plugin [184] was used instead to replicate the view control while providing more access to the view control data.
5.6
Summary
In order to host the two-stage experiments designed to generate data for utility evaluation, a new research flight simulator had to be designed. The goal was to have a simulator system representative of its low-fidelity class so that experimental results could be validated against other systems qualified to a similar standard. To ensure a transparent and accountable system design, a top-down systems engi- neering process approach was used. This process meant that several iterations of requirements and functional analysis steps were completed to synthesize the final design.
The system requirements were also influenced by the fact that the simulator also had to support other projects. The inputs gathered to determine the total set of requirements were obtained from collaborating with AVRRC colleagues. This re- sulted in the identification of two critical requirements: fixed-based simulator and minimum frame rate of 60 Hz. Driving requirements were the use of commercial- off-the-shelf components to reduce cost and representative, reconfigurable flight
controls and displays. Matching the low-fidelity class of simulators, the key re- quirement was a small physical footprint and mobility for quick transport.
The final design of the simulator system architecture is based on a twin-station configuration running X-Plane: an experiment station where the participant is ac- tively controlling the aircraft and a control station where the observer supervises the experiment. The main hardware features are a fighter-type cockpit shell hous- ing three computer displays for the experiment station powered by a single PC. The control station consisted of two PCs driving the replicate cockpit displays for monitoring the experiment and the instructor operator console for manipulating the experiment settings. The triple projectors uses manual, software calibration to warp and blend a single, large display. Head-tracking for amplified head rotations was done using a single, infrared TrackIR camera and markers.
Results
As Figure6.1illustrates: after carrying out the experiments designed in Chapter4
with the simulator system built in Chapter 5, this chapter presents the results of both experimental evaluations after performing the statistical analysis methods described in Section4.4. The experiments and the obtained results provide insight into benefits obtained with augmentation compared to a standard setup and serves as a launch platform for further experiments with augmentation on larger displays. This forms the basis leading to a full discussion of the results and its implications in Chapter 7.
Figure 6.1: Chapter6 in thesis roadmap
6.1
Experiment 1: Base-to-Final Circuit
E
leven male participants were recruited from the university for this exper- iment. All were current engineering students with a background or keen interest in aviation. The mean age was 28 years, with a standard deviation of seven years. Although some participants had logged flight time as a pilot, all were briefly assessed prior to the experiment on basic flying manoeuvres such as turns, descents and landings so that they could be considered equally capable of carry- ing out the experiment task. None of the participants had prior experience with amplified head rotation applications.The participants generated a total of (11 × 12 =)132 measurement runs. With a single group of participants and one independent variable (DISP), a one-way ANOVA was utilized for statistical analysis where the assumptions of normality and homogeneity of variance was met. Some dependent variables were not nor- mally distributed and had heterogeneity of variance, of which skewness could not be corrected through transformation, therefore a non-parametric Kruskal-Wallis test was applied instead.