In the recent years, with the explosive growth of the computational power, the machine learning algorithms and the so called deep learning keep gathering more ground, whereby the reserachers are becoming able to succeed on more and more domain of technical problems. These algorithms nowadays can solve complex problems on a human level or even beyond, like image recognition or playing so difficult games as Go.
The topic of this bachelor thesis is the possible application of these algorithms in the field of control engineering, within it, the theoretical realisation of the control of a racing car. For the implementation, I used reinforcement learning based algorithms as well as deep neural networks.
While solving the task, first, using a simple example I compared the task relevant toolboxes of Matlab and Python. Then, I designed and implemented the simulator framework later used for testing the control algorithms. On this grafical user interface, the necessary parameters of the environment, the car and the contol algorithm can be set, as well as the progress of the learning can be observed. Finally, I implemented and compared the performance of some task relevant deep reinforcement learning based algorithms.