Data-driven Understanding of User Interaction with Human Machine Interface (HMI) in the Automotive Industry

Abstract

User behavior evaluation within the automotive field has traditionally been based on qualitative methods like interviews and surveys. However, the improved data availability and a stronger focus on data utilization make this approach change fast towards the data-driven evaluation since decision-making in development needs an evidence-based approach. Numerous studies of user interaction based on quanti tative methods like data logging have been performed. Most of them are focused on external ECU communication (CAN, LIN, Ethernet, etc.) or video recordings. To fully understand all user interaction steps, data collected directly from the HMI needs to be studied. Therefore, as a first step, semi-structured interviews with HMI Engineers at Volvo GTT were conducted to determine relevant logging points. Afterward, a user interaction logging setup was developed for an Instru ment Cluster in trucks, based on internal ECU data. Data logged includes usage of Instrument Views, Focus Shift, Gauges and a menu application. In designing the logging, parts of the HMI have been modelled using Discrete Event Systems theory to identify the system behavior. Moreover, the analysis has been made to determine the most suitable level of the software to extract wanted data. Finally, driver card information has proven to be a beneficial way to identify unique users within the truck industry. Logging is done in the form of DOIDs, a type of parameters with static definitions of the data structure. However, for the future, a more dynamic high-frequency logging setup is needed to connect logged user data to driving conditions and system settings. As the user interaction logging has been verified successfully, a mixed methods approach is proposed once real customer data is available. Logged data standalone will not explain patterns found in the data, but a more complete picture can be given when combining quantitative and qualitative methods.