Driver re-engagement in autonomous driving by means of HMI adapted to human activity
There are several challenges that need to be researched to create safe transition strategies in future vehicles. For example, it takes time to re-engage in the driving task after being engaged in another activity. Re-engagement to manual driving has to be safe and provide a positive experience. More knowledge is therefore needed about for example HMI linked to the driver's activity and condition. In addition, we need more systematic knowledge about what is perceived as a good or less positive HMI. Knowledge of how machine learning and UX can be combined in vehicles is another important area for the Re-engage project. The project builds further on DRAMA, a former SAFER project that developed methods for supervising the driver and passengers in the car to give them the best possible support in different situations. The uniqueness of this project is that it combines UX and Machine learning in the same project.
Machine learning supports the project
The main task of the project is to explore, implement and demonstrate solutions for driver support when regaining control again when the vehicle has been active in automated mode. The researchers will work with recognition of driver activities, such as taking a power nap and reading a book. In this part of the project, technology for camera-based machine learning will be used. The knowledge can then be used to create safe and comfortable user experiences that are tailored to the driver's activity. The research team also wants to explore new interaction patterns and work on how interaction can be adapted to activity. Studies will be conducted to examine what different drivers think about the experience of taking back control. Example of research questions that the researchers will look into:
• What makes AD valuable for people in the car?
• What activities can be expected in AD cars?
• How can these activities be detected and classified?
• How can we design human-vehicle re-engagement interactions tailored to specific activities?
• What does the design space look like for activity tailored HMI?