Thursday, December 22, 2022

Machine learning based trajectory planning in cloud architectures

Topic and Goal of the Thesis

Automated vehicles require vast amounts of data gathered by different sensors usually directly mounted to the vehicle itself. Many implemented functions use machine learning algorithms that were trained before their implementation into the vehicle. The goal of this thesis is the implementation of a cloud-based functionality for the determination of trajectory suggestions for connected and automated vehicles in the so-called collective memory. For this purpose, an already existing approach is tested and optimized. Alternatively, a new approach can also be designed and implemented. The work is part of the UNICARagil project, in which new architectures for automated vehicles are developed and demonstrated in four prototypes.

Working Points

• Familiarization with AI toolchains and UNICARagil structures
• Evaluation of the current approach for data-driven trajectory planning as well as optimization of the network architecture
• Evaluation of the implemented architecture

Requirements

• Good English and German language skills
• Reliability, commitment and enjoyment of working independently
• Experience with C++ and ROS
• Experience with Machine Learning is an advantage

Note: Please attach brief resume and grade summary.