| Summary: | 碩士 === 國立交通大學 === 工業工程與管理系所 === 107 === Car manufacturers and technology companies are dedicating to developing self-driving
technique. No matter how good the technique does, there is a limitation and the possibility of failure. Therefore, users must take the responsibility of monitoring and take over the control when the system falls into the emergency. However, the robustness and reliability of the system will modify the user's trust and the dull and dry monitoring task will cause the attention transfer to other tasks. These will lead to lacking the vigilance and situation awareness causing the critical incidents. The aim of the research is to investigate the display design that could reveal the status of a self-driving system transparently. Which could make users aware of the situation immediately and take-over on time. We utilize an ecological interface design (EID) approach to design the pattern of information and expect to present the status clearly and helping out the take-over the process. This research recruited 34 participants randomly assigned to two different information pattern
interface groups. Participants will be asked to monitor a simulation with the active autonomous driving system while distracting by secondary tasks. Participants were told that they could stop the simulation by clicking the button when they felt the situation was out of the system’s capability. At the moment they stop the simulation, the time will be recorded and the screen will switch to the questions of situation awareness(SA) that they would need to answer. According to their respond giving them SA score, and classifying driver’s trust level by their reaction time (RT) distribution into three groups: over-trust, moderate, and under-trust group. The result shows that while distraction, under-trust and moderate group will rely on traditional
ways (situation and instrument cluster) to capture the information elements of the situation so these two groups had better perception (SA level 1) performance. Over-trust drivers had a tendency to use the transparent EID to predict future status so they had better performance on projection (SA level 3). By interviewing participants, under and moderate trust group rely on traditional ways of gathering information to understand the situation but their information capacity are overloaded that led to poor prediction performance. Over-trust group rely on transparent EID rather than collecting low-level detail information so they had better high-level prediction performance. To conclude, designers should consider the driver’s trust during designing the in-vehicle
interface that has a great influence on the usage of the interface. We can embed dynamic trust calibration mechanism during the use of an autonomous system and according to the reaction time giving salient information feedback. Balancing driver’s workload and their situation awareness that let the drivers have better coping take-over process in order to improve safety could be future research issues.
|
|---|
