Uploaded August 2026 | Updated September 2026, 5 hours ago
This is the project video for Etienne Bratschi's bachelor's thesis in the computer science department at the Bern University of Applied Sciences, Spring 2026.
Drive-by-wire systems enable vehicle control through specialized human-machine interfaces (HMIs), offering increased accessibility for drivers with physical disabilities. However, operating these HMIs requires substantial training due to their limited input range and different driving characteristics compared to conventional steering wheels and pedals. To address this challenge, this thesis presents joysteerVR, a virtual reality driving simulator developed for Bozzio AG’s
Joysteer drive-by-wire platform.
The thesis extends an existing Godot-based simulator by improving simulation accuracy, expanding HMI support, enabling runtime HMI hot-swapping, implementing a certification training environment, enhancing virtual reality interaction, and optimizing overall system performance. In addition, a more comprehensive training framework was implemented to calibrate vehicle simulation parameters using real-world driving data collected from a VW T6 equipped with the joysteer system.
Vehicle behavior was optimized using a differential evolution algorithm trained on multiple driving scenarios, including track driving, acceleration, braking, and cornering exercises. The new training architecture significantly reduced optimization time by greatly accelerating iteration time and introducing in-application iteration reloading. Experimental results demonstrate substantial improvements in trajectory accuracy compared to previous work, reducing average positional error from approximately 175 m to 66 m in the primary evaluation scenario. Furthermore, a new certification map containing realistic driving exercises was created to support structured driver training and performance assessment.
This is the project video for Etienne Bratschi's bachelor's thesis in the computer science department at the Bern University of Applied Sciences, Spring 2026.
Drive-by-wire systems enable vehicle control through specialized human-machine interfaces (HMIs), offering increased accessibility for drivers with physical disabilities. However, operating these HMIs requires substantial training due to their limited input range and different driving characteristics compared to conventional steering wheels and pedals. To address this challenge, this thesis presents joysteerVR, a virtual reality driving simulator developed for Bozzio AG’s
Joysteer drive-by-wire platform.
The thesis extends an existing Godot-based simulator by improving simulation accuracy, expanding HMI support, enabling runtime HMI hot-swapping, implementing a certification training environment, enhancing virtual reality interaction, and optimizing overall system performance. In addition, a more comprehensive training framework was implemented to calibrate vehicle simulation parameters using real-world driving data collected from a VW T6 equipped with the joysteer system.
Vehicle behavior was optimized using a differential evolution algorithm trained on multiple driving scenarios, including track driving, acceleration, braking, and cornering exercises. The new training architecture significantly reduced optimization time by greatly accelerating iteration time and introducing in-application iteration reloading. Experimental results demonstrate substantial improvements in trajectory accuracy compared to previous work, reducing average positional error from approximately 175 m to 66 m in the primary evaluation scenario. Furthermore, a new certification map containing realistic driving exercises was created to support structured driver training and performance assessment.
