라즈베리파이 기반 자율주행RC카의 딥러닝모델 비교 연구
A Comparative Study of Deep Learning Models for an Raspberry Pi based Autonomous RC Car
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
The Raspberry Pi, characterized by its low power consumption, affordability, and compact size, is highly suitable for developing embedded vehicular systems for autonomous driving research. Furthermore, its high compatibility with various sensors and communication modules enables flexible implementation of autonomous driving functionalities. In this study, a Raspberry Pi 4-based platform was utilized to compare steering angle prediction performance between the nVidia CNN model and the EfficientNet-Lite model. Autonomous driving was successfully achieved on a custom-built complex track without lane deviation. Additionally, by converting floating-point tensors to integers, the model maintained its performance while reducing computation time by up to one-tenth.
Abstract
The Raspberry Pi, characterized by its low power consumption, affordability, and compact size, is highly suitable for developing embedded vehicular systems for autonomous driving research. Furthermore, its high compatibility with various sensors and communication modules enables flexible implementation of autonomous driving functionalities. In this study, a Raspberry Pi 4-based platform was utilized to compare steering angle prediction performance between the nVidia CNN model and the EfficientNet-Lite model. Autonomous driving was successfully achieved on a custom-built complex track without lane deviation. Additionally, by converting floating-point tensors to integers, the model maintained its performance while reducing computation time by up to one-tenth.