Autonomous driving and machine learning with TurtleBot3 Waffle Pi mobile rover

Abstract

This project explores autonomous driving and machine learning with the TurtleBot3 Waffle Pi mobile rover in a custom-built racing rig. The project aims to develop an autonomous driving vehicle research and educational platform where the TurtleBot3 navigates the racing rig while clearing obstacles such as traffic lights, intersections, parking, construction work, and tunnels. Various methods are employed to achieve the objective, including camera calibration, image processing, feature detection, SLAM-based navigation, and reinforcement learning. The robot's performance is assessed in both simulated and real-world environments.

In the simulated environment, the robot successfully completes the course but encounters issues such as cutting corners and occasional difficulties generating a path through the tunnel. In the real environment, the robot can detect and follow the lane and clear obstacles, traffic lights, and tunnels but struggles with traffic sign detection, affecting its ability to start different missions.

The paper provides insights into the capabilities and limitations of the TurtleBot3 for autonomous navigation in complex environments and identifies areas for further development and optimization. This research lays the foundation for future work in autonomous driving and mobile robotics.