This is our team’s most recent project, focused on the autonomous human-machine interaction system for prediction and automated execution of certain comfort functions of a car. For example, let’s assume a driver, who is regularly driving his car on a highway. In summer, they would always turn on air-condition in the car, and listen to music using the onboard infotainment system. During winter though, the situation would change. Because of lower temperatures, and increased risk of traffic accidents on the way they would turn up the temperature in the car and tune the onboard radio to the news. Since doing such mundane task every single time the driver uses their car can be inconvenient, the basic idea of this project is to provide a platform for AI-driven automatization of such tasks. The system is based on collection of communication data within the onboard data buses and environmental sensors, feeding the data into third-party machine learning models, and in case an action (e.g., turning on air-condition) is required, it is proposed to the user. If they accept proposed action, it is executed afterwards.

The developed system consists of various subsystems. Data acquisition and actor subsystems are responsible for acquiring the vehicle operational data from the onboard automotive buses and executing required actions on car subsystems. The HMI subsystem is responsible for presenting proposed actions to the car user and obtaining their input. The decision subsystem evaluates the operational data obtained by the data acquisition subsystem and outputs a decision about action to be automatically executed on the car by utilizing third-party machine learning models. Because of the variety of involved subsystems, this project offers a wide range of tasks to participate – from PCB and FPGA design, through low-level programming, to high-level, cloud-native programming tasks. Currently, the system is incorporated in Škoda Kodiaq and Škoda Enyaq Coupé RS iV thanks to cooperation with Škoda Auto.