Human-Computer Interaction Institute
School of Computer Science, Carnegie Mellon University


Designing Automatically Generated
Perceptually Optimized Displays

Joonhwan Lee

February 2008

Ph.D. Thesis


Keywords: Maps, in-car navigation systems, visualization, human attention, perceptual optimization, dynamic displays, situationally optimized interaction

Most current user interfaces do not carefully consider particular "situations" or contexts of use, thus providing information to the user with the same demands of attention no matter what the user's attentional state is. This can result in serious breakdowns in communication between the user and the system in some situations, and is witnessed very often in our daily lives.

The driving context is a good example of a need for situationally appropriate user interaction. Drivers often need assistance when navigating an unfamiliar route. But the displays created by location-based software such as GPS mapping applications are often not straightforward enough to use in the context of driving. Information is crowded and overloaded on the display. Critical information is designed and presented in a way that slows down the rate of uptake, interfering with the process of learning and remembering the route, encoding the information in memory, and making decisions at critical points. To address the attentional demands of reading a map while driving, we developed the Maps Optimized for Vehicular Environments (MOVE) in-car navigation display, which provides situationally appropriate navigation information to the driver through optimization of map information.

In this dissertation, we describe the iterative design and evaluation process that shaped the MOVE system. We describe the map reading and navigation studies that led to early designs for our system. We present a study on visual search tasks that refined the renditions used for the system. Then, we present our second study on the effectiveness of several variations of a perceptually optimized route map visualization with a desktop steering system. The result of this study shows that MOVE's perceptually optimized navigation information can reduce the driver's perceptual load significantly. Our laboratory experiment shows that the total map display fixation time was decreased six-fold, and the number of glances to interpret the map display were decreased about threefold, when comparing the contextually optimized display to a static display.

We then describe the process of implementing the MOVE system, bridging the design and research we have taken from our preliminary studies. The implementation process presents the following steps: First, the implementation of the Road Layout process generates the entire route as simply as possible, while making the important portions of the route segment salient. Second, the Rendition Selection and Rendition Scoring process selects the appropriate forms of map features to lower the driver's attention to the display by reducing the overall amount of information presented. Third, the Final Placement Tuning process uses an intervention technique to prevent possible conflicts and clutter within the selected renditions when presented on the display. Then we present our final evaluation study of the perceptually optimized displays in the context of real driving. The result of this study also shows that the MOVE's perceptually optimized navigation information can reduce the driver's perceptual load significantly.

Finally, we present a summary of contributions and plans for future work.

195 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by