Abstract

Today, it is known that a through understanding of the end-user is the most valuable information to improve design, training, maintenance and assembly/disassembly processes for products or systems. The most widely used method for this purpose, user experiments, requires a product prototype, a test environment and a researcher to watch and collect the data. This results in a procedure that is time consuming and limited by geographical constraints. Although remote testing is being used today by recording or transmitting the test situations, these still depend on audio or video data that require a researcher to watch and analyze. Therefore there is a need for tools that collect and analyze user data automatically from actual interactions of the user with the product. This thesis proposes an approach to achieving that goal. The approach proposed integrates existing virtual reality technology with a four-phase analysis method that uses techniques from data mining and human-computer interaction researches. In our approach, we first cluster subjects into clusters such that subjects with similar performance are placed into the same cluster. Then, for each cluster, we extract paths that have been followed by users during their interaction with the product. Finally we demonstrate these paths together with the statistical results obtained for a cluster. As result the researchers see if the users are following the paths that they were expected to follow, what are the common paths followed by users with a better or relatively poor performance, which aspects of the design can be changed to direct the users to better paths or how can the users be trained better to follow desired paths which leads to easiest, safest, and most effective working conditions. The thesis describes the model constructed for the approach and then presents an example application: simulation study in the evaluation of a simple assembly process. Initial results support the usefulness of method as an automated tool to detect problems.