Abstract:

The paper presents the design of the laboratory works for the subject of Autonomous Vehicles, newly introduced at the master program of Research and Development in Vehicle Engineering from the Faculty of Transports of the University POLITEHNICA of Bucharest, Romania. The particularities of this subject are, beside the momentary lack of dedicated gear, the prerequisites of electronics and computer science for graduates in mechanical engineering, with limited skills in these subjects. For this reasons, the lectures will be attended by laboratory works where the students will study the fundamental concepts of self-driving cars on reduced scale model cars, controlled by Raspberry Pi Single-Board Computers. They will familiarize themselves in the first laboratory works with the Python programming language and with the Raspbian operating system. They will learn later to use lighting elements (LEDs), sensors (ultrasound, infrared, laser), moving elements (DC-motors and servo-motors) and power elements (batteries and drivers for the motors). On this foundation, they can study aspects like sensor fusion and can build small-scale cars which can run, detect obstacles and even perform the parking maneuvers. On the second part of the labs, the students are learning how to use a video camera attached to the Raspberry Pi, to do image processing and to stream the video and other informations. They will be able now to perform real-time Advanced Driver-Assistance Systems (ADAS) functions like lane centering. The last part deals with working with Deep Learning on the Raspberry Pi using Coral, the Google’s Edge Tensor Processing Unit accelerator. Some of the new results are pedestrian detection, traffic sign recognition and they are attended by the improvement of the previous ones, like adaptive cruise control. The laboratory introduces the students into the atmosphere of automated control and machine learning and prepare them for the vehicles of the future, the autonomous ones.

Keywords:

Autonomous vehicle, Self-driving car, Laboratory, Raspberry Pi, Python, Deep Learning, TensorFlow.