Master’s Thesis Student in Uncertainty Quantification and Reliability Assessment of Virtual AI Sensors for Chassis Systems

Mercedes-Benz Group AG · Sindelfingen

ThesisGerman needed: fluent

Still open on 10 Nov

You apply on the company's own website.

Key facts

Language needed: German and English
Subject you study: Engineering, Computer Science, Physics, Mathematics, Applied Mathematics

Study current methods for uncertainty quantification, such as conformal prediction, quantile regression and Bayesian methods, as well as out-of-distribution (OOD) and drift detection
Develop a model-independent concept that combines prediction intervals with reliability assessment and can be applied to different regression models
Distinguish between methods requiring reference data during development and methods working without labels in customer vehicles
Implement and validate the approach in the existing Python framework using real measurement data and defined changes in vehicle conditions
Assess usefulness, computing requirements and memory requirements for possible future embedded implementation
Optional: Apply uncertainty estimates to measures relevant to structural durability and develop an embedded prototype
Document the results and prepare them for scientific presentation

Studying engineering, computer science, physics, mathematics or applied mathematics, or a comparable subject
Strong enthusiasm for machine learning and data science, ideally with some practical experience in this area
Experience in Python programming and data analysis
Strong analytical skills and a creative, independent approach to work
Enjoy scientific research
Teamwork skills and commitment
Confident written and spoken German and English
Bonus: Knowledge of uncertainty quantification and OOD, drift or anomaly detection
Bonus: Knowledge of chassis engineering or structural durability

Opportunity to contribute to a forward-looking, research-focused topic in vehicle development
Insights into the structural durability of modern chassis systems
Work in an experienced team with people from different disciplines
Support from subject experts throughout the thesis
Use of real measurement data from vehicle development

Study current methods for uncertainty quantification, such as conformal prediction, quantile regression and Bayesian methods, as well as out-of-distribution (OOD) and drift detection
Develop a model-independent concept that combines prediction intervals with reliability assessment and can be applied to different regression models
Distinguish between methods requiring reference data during development and methods working without labels in customer vehicles
Implement and validate the approach in the existing Python framework using real measurement data and defined changes in vehicle conditions
Assess usefulness, computing requirements and memory requirements for possible future embedded implementation
Optional: Apply uncertainty estimates to measures relevant to structural durability and develop an embedded prototype
Document the results and prepare them for scientific presentation

Studying engineering, computer science, physics, mathematics or applied mathematics, or a comparable subject
Strong enthusiasm for machine learning and data science, ideally with some practical experience in this area
Experience in Python programming and data analysis
Strong analytical skills and a creative, independent approach to work
Enjoy scientific research
Teamwork skills and commitment
Confident written and spoken German and English
Bonus: Knowledge of uncertainty quantification and OOD, drift or anomaly detection
Bonus: Knowledge of chassis engineering or structural durability

Opportunity to contribute to a forward-looking, research-focused topic in vehicle development
Insights into the structural durability of modern chassis systems
Work in an experienced team with people from different disciplines
Support from subject experts throughout the thesis
Use of real measurement data from vehicle development