Nowadays neural networks are more and more popular due to the advancements of the past few years. In many areas neural networks manage to achieve better results than the humans or the classical algorithms.
Enhancing passive and active safety is one of the most important principles of the automotive industry. ADAS (Advanced Driver Assistance Systems) systems, like the ACC (Adaptive Cruise Control) are designed to enhance active safety. Besides it’s comfort functions, the ACC was designed to reduce the possibility of rear-end collisions.
The aim of this thesis was to find out if the adaptive cruise control functionality can be realized based on neural networks. In order to achieve this, I had to get a comprehensive knowledge of the area of neural networks, I needed to create training datasets based on public road measurements, than I had to train neural networks with different parameters. I evaluated the results of the trainings, than I developed an application, which makes possible to test the system. The application reads the inputs from a vehicle’s CAN bus, evaluates the neural network with these inputs and writes back the output to the vehicle’s CAN bus. The system was tested with simulation and prepared to vehicle tests.