Abstract:Recent advances in computer and information technology have led to the feasibility of sophisticated automotive displays that can help drivers in choosing efficient routes. However, there is a concern that providing guidance information in the vehicle could result in distraction from the driving task. This dissertation presents a comprehensive effort in conducting simulator experiments to study the effect of in-vehicle route guidance systems on driving performance. Three experiments were conducted as part of this research. The first two experiments tested a paper map along with a set of electronic visual and audio route guidance systems. Each subject drove through a simulation environment that included pedestrians, stop signs, traffic signals, crossing and turning traffic, and additionally in experiment 1, a lead vehicle. Results indicated that subjects had the highest workload, worst perception ratings, largest number of navigation errors, highest reaction times, and lowest speed when using the paper map. By comparison, audio systems were associated with the lowest workload, best perception ratings, lowest number of navigation errors, lowest reaction times, and highest speed. The location of the visual display had a significant effect on driving performance. Subjects responded faster to external events with the heads-up turn-by-turn display compared to the identically designed heads-down turn-by-turn display. In spite of its complexity, the heads-down electronic map was received quite favorably and associated with lower reaction times compared to the heads-down turn-by-turn display. Experiment 3 (conducted using a table top simulator) indicated the following: synthesized speech was less preferred to recorded speech; turn-by-turn displays that included street names, symbolically provided distance information and turn intersection geometry were preferred; a simplified electronic map (with fewer street names) was associated with improved tracking task performance compared to a more complex map. This dissertation also illustrates several approaches for modeling driving performance. Reaction time data has traditionally been modeled using ordinary linear regression models. These commonly applied models are unsuitable for censored data, necessitating the application of new statistical approaches. In this dissertation, tobit models were developed for analyzing censored reaction time data.

Note:Ph.D. Dissertation