A self-driving system developed with Linux & ROS on an NVIDIA Jetson TX2, to demonstrate dynamic inductive charging from the road. The sensor package consists of a lidar and camera. Hector SLAM is used for mapping and localization. Pure Pursuit is used to follow a manually recorded path. Machine-learning and behavior cloning is used to make the vehicle drive with camera only. Recorded camera images and steering input were used as training data to a deep neural network. The truck is controlled with PWM-signals from a Teensy 3.2 microcontroller. There are four different driving modes; manual, record new path, lidar/SLAM-mode, and AI-mode. The truck has an onboard induction charger system that talks with the TX2 through CAN-bus. There is also a custom GUI to monitor the states of the truck and battery.

Link to paper: https://ntnuopen.ntnu.no/ntnu-xmlui/handle/11250/2625670
Github: https://github.com/joneivind/Self-Driving-Truck

Master Thesis in Cybernetics & Robotics by Jon Eivind Stranden at the Norwegian University of Science and Technology (NTNU) in cooperation with SINTEF Energy Research, 2019.
Source: joneivinds YouTube channel