Efficient Self-Driving: A Study on the Behavioral Cloning based Autonomous Navigation

In the past five years, Engineering Physics students at the University of Central Arkansas investigated autonomous robots that utilize behavioral cloning. The current team, UCA Jetson, presented an advanced self-driving robotic platform. Key contributions include redesigning the mechanical and software systems. The new mechanical system consists of a Jetson Orin NX Development Board, a Raspberry Pi Pico W, an Intel RealSense D455 Camera, a brushless motor, and an electronic speed controller (ESC). The software improvements incorporate EfficientNet, a deep learning model, along with new integrated mechanical parts, while maintaining the software’s structure. These upgrades enhance the robot’s image-based decision-making for autonomous navigation. The results show that the upgraded robot demonstrated superior autonomous performance on various race tracks during the experiment, while outperforming its predecessor. This research showcases a successful integration of hardware and software upgrades on a smaller-scale self-driving robot.