ScholarWorks@숙명여자대학교

초록

As artificial intelligence continues to be adopted across various industries, it is bringing new vitality to the mobility sector. Among these advancements, smart mobility has emerged as a promising field capable of improving quality of life and creating new industries. Autonomous vehicles are considered a core technology of smart mobility, driving research into autonomous driving algorithms using model vehicles. However, platforms that allow researchers to study and apply the principles and concepts of autonomous driving from start to finish are still limited. In this study, we developed an autonomous vehicle using the Raspberry Pi 4, a representative microcomputer, and applied a deep learning model based on Nvidia's CNN. The autonomous vehicle utilizes a 2-channel DC motor connected to the GPIO pins of the Raspberry Pi. Steering angles are calculated using a mouse, and the speed of the left and right motors is adjusted accordingly to demonstrate autonomous driving on a track. Images captured during the process were uploaded to Google Colab, where a T4 GPU was used to generate the model. The resulting model was then transferred back to the Raspberry Pi to perform autonomous driving. A track was designed to allow the examination of various driving characteristics, and the model was trained using the captured images to test autonomous driving. The results showed that the vehicle successfully completed all routes without deviating from the track at any point.

키워드