Model training of a simulated self-driving vehicle using an evolution-based neural network approach

Type of thesis: Bachelorarbeit / location: Leipzig / Status of thesis: Theses in progress

In the simulation, a control model for a self-driving vehicle is to be trained and evaluated with an artificial neural network. In order to evaluate the quality of the control model, the time from the start to the finish line is measured in a course. During the course, it has to navigate around red and blue obstacles. As a gold standard, an optimal route by human standards through the course is designed, implemented and the time is measured. An evolutionary algorithm will be used to train the neural network.

 

Counterpart

Dr.
Thomas Burghardt

Service and Transfer Center, Living Lab

Universität Leipzig

Production and Logistics, Industry 4.0, Connectivity, IoT Platform, Programming

Tobias Jagla

Universität Leipzig

Swarm Intelligence, Robotics, Machine Learning

TU
Universität
Max
Leibnitz-Institut
Helmholtz
Hemholtz
Institut
Fraunhofer-Institut
Fraunhofer-Institut
Max-Planck-Institut
Institute
Max-Plank-Institut