Modern vehicle control algorithms

OData support
Supervisor:
Kiss Domokos
Department of Automation and Applied Informatics

Nowadays, there are more and more automotive companies starting to work on new solutions to replace human drivers in commercial vehicles with a reliable autonomous system. The number of car accidents are increasing continuously due to the increasing number of vehicles on the roads. This new research project intends to develop a safer transporting environment since mostly the drivers are causing the accidents. Additionally, there are more advantages of autonomous driving: all the passengers can do their average daily routine while travelling; the parking might no longer be a responsibility of the driver, because the car might be able to search for parking slots itself; by creating a communication interface between the vehicles, detected threats on the roads can be shared among the cars; etc. Therefore, there are many reasons to research this field.

In the autonomous driving subject, my task was to investigate path tracking methods, focusing on non-linear MPC algorithms. It is assumed throughout the project that the steering actuator system is a working element of the vehicle having known parameters, the vehicle state is measurable by sensors and the reference path is available.

In my thesis, firstly I examine the possible simulation elements, the solvers and vehicle models to later use them in the MPC tracking algorithm. Secondly, to create an optimal control algorithm I introduce two possible optimization methods, the gradient descent method and the Nelder--Mead method and explain the main elements, difficulties and implementation possibilities of these. By combining the methods with the vehicle models I created Simulink control algorithms and presented the results. It is examined which method can be implemented in a real time application, furthermore, the gradient descent method with the kinematic bicycle model is uploaded to a dSPACE MicroAutoBox in a testing vehicle and results of field tests are presented.

Additionally, it is intended to show that the presented optimization techniques are appropriate for other subtasks of autonomous driving as well. Therefore, possible reference path optimization algorithms are presented briefly, at the end of the thesis.

I wrote this thesis as an intern at thyssenkrupp Components Technology Hungary Kft.

Downloads

Please sign in to download the files of this thesis.