Promotion von Daniel Baumann mit "magna cum laude" ausgezeichnet!

Daniel Baumann successfully completed his doctorate on November 10, 2026. His dissertation, titled “Reconfiguration of Learning, Cloud-Based Vehicle Functions: The Case of Air Conditioning in City Buses,” was awarded the distinction of magna cum laude. The doctorate was supervised by Prof. Dr.-Ing. Eric Sax. The co-supervisor was Prof. Dr.-Ing. Clemens Gühmann from the Technical University of Berlin.
Through his research, Daniel Baumann is making an important contribution to the development of intelligent, data-driven vehicle functions. His academic work focuses on the digitalization and connectivity of vehicles, cloud-based systems, data-driven methods, and machine learning techniques for optimizing complex technical systems.
Abstract of the Dissertation
The increasing electrification and digitalization of vehicles is shifting the focus of innovation from hardware to software. As part of this development, cloud-based vehicle functions are gaining importance, as they enable the centralized aggregation and processing of fleet data.
This dissertation develops a concept for the reconfiguration of learning, cloud-based vehicle functions during operation, based on the Cross Industry Standard Process for Data Mining (CRISP-DM). Vehicles are initially delivered with a general data-driven function, whose functionality is reconfigured contextually during operation through the continuous analysis of fleet data.
A prototypically implemented use case addresses the temperature setpoint specification for the heating, ventilation, and air conditioning (HVAC) system of city buses, as this function offers high energy-saving potential with moderate time constraints. The optimization of the temperature setpoint is based on predicted disturbance variables, with solar radiation and outside temperature derived from weather forecasts, and vehicle speed, door status, and passenger count predicted using data-driven methods.
To model heterogeneous operating conditions for this prediction, a Mixture-of-Experts (MoE) architecture is used, in which context-specific expert models are derived from fleet data. In contrast to classical MoE approaches with a static model structure, the concept provides for a flexible, data-driven number of experts. A gating function controls the context-dependent selection of experts, while reconfiguration is based on newly available fleet data and enables adaptive adjustment to changing contextual conditions.
The evaluation is based on the Zurich Transit Bus dataset (ZTBus dataset) from the regular service of a city bus fleet in Zurich over a period of 3.7 years and comprises 1,409 trips. It shows that cloud-based temperature setpoint control using a MoE architecture reduces the HVAC system’s energy consumption by 11.58% compared to static setpoint control based on the characteristic curve of the Association of German Transport Companies (VDV). Of this reduction, 1.46 percentage points are attributable to the context-sensitive MoE approach compared to a general model. At the same time, the duration of deviations from the comfort characteristic map according to VDV Publication 236 is reduced by a factor of 3.8. The transferability of the approach is confirmed by a 6.34% savings at a second location in Hamburg with 733 cycles.
Daniel Baumann’s dissertation thus demonstrates that a context-dependent, reconfigurable, cloud-based vehicle function provides a suitable foundation for the energy-efficient and adaptive implementation of data-driven functions in vehicle operation.
The ITIV warmly congratulates Daniel Baumann on this successful completion and on his dissertation being awarded the grade of magna cum laude.


