Low Visual Fidelity in Driving Simulation
Tutor / director / evaluatorde Winter, Joost
Document typeBachelor thesis
Rights accessRestricted access - author's decision
This thesis consists of an initial survey of the research literature in driving simulation and, specifically about the influence of visual fidelity on driving performance and behaviour. The main section comprises the preparation, execution, and analysis of an experiment with the Green Dino fixed-base driving simulator within the 3mE faculty at TU Delft. This experiment evaluated three levels of visual fidelity: (1) a realistic high level, (2) a medium level with colours and horizon, but without textures and objects, and (3) a low level like the medium one but without colours and without horizon. Participants drove along a rural road without speedometer. Speed and driving performance data was collected along with eye tracking information. A within-subject statistical analysis was performed to determine the differences and resemblances between the visual levels and further conclusions are drawn with assessment provided by the eye tracking data. Results showed small but significant differences in all the driving behaviour and performance measures between the highest level and the two lower ones. However, no differences were found for those measures between the medium and low levels, suggesting that decreasing the optical flow or removing the textures has a stronger influence on driving behaviour and performance than removing the colours and the horizon. Focusing on specific measures, the steering activity was higher, as well as the lane keeping precision, in the high level than in the medium and low ones. Driving speed was slower in the medium and low levels than in the high one. Drivers tended to look lower as fidelity level decreased and, since the high visual fidelity level had roadside objects and textures, the gaze variance was larger in that level.
|PFC Arnau Ramon Setembre 2013.pdf||Report||2.642Mb||Restricted access|