Thesis - Machine Learning for the virtual testing of autonomous driving vehicles

Master Thesis

Automotive companies are intensively developing automated driving functions and vehicles. Unfortunately, we have already seen severe road accidents involving automated vehicles. Thus, the safety and reliability of autonomous driving vehicles (ADV) need to be tested and verified under different environmental and traffic conditions. But, how to test ADV in a safe way for all critical situations?

A naïve approach would be virtual testing of all possible parameter and scenario combinations that influence the ADV sense, plan, and action system. Unfortunately, such testing with a full factorial variation of the parameters is not feasible. In AVL, we are developing methods for the efficient identification of critical scenarios.

TASK:

• Performance comparison of the Machine Learning methods developed at AVL
• Identification of the limitations of the Machine Learning methods developed at AVL
• Analysing the methods by using cognitive testing methods available in the literature on a publicly available AD stack like Autowave and Apollo

REQUIREMENTS:

• Good programming skills in Python or C++
• Knowledge of Machine Learning
• Good knowledge of German and English
• Knowledge of optimization methods is beneficial
• For this thesis is your presence at our headquarter in Graz required!

PREFERRED FIELD OF STUDY:

• Computer Science
• Telematics
• Physics
• Electrical Engineering

The successful completion of the thesis is remunerated with a one-time fee of EUR €3.500,00 before tax.

You don't want to write your final thesis just for the books, then explore the mobility of the future together with us! Maybe you will be a part of it soon!