@article{M53EEAB35,
title = "Research Trends Focused on End-to-End Learning Technologies for Autonomous Vehicles",
journal = "The Journal of Korean Institute of Communications and Information Sciences",
year = "2024",
issn = "1226-4717",
doi = "10.7840/kics.2024.49.11.1614",
author = "Jeong-hwan Choi, Yechan Park, Woomin Jun, Sungjin Lee",
keywords = "Autonomous Driving, End-to-End Learning, Modular Architecture, Perception, Planning, Control",
abstract = "As autonomous driving technology is researched as a core technology in the future automotive industry, recent advancements in related sensors, components, platforms, and artificial intelligence technologies have led to rapid progress in this field. This paper focuses on end-to-end learning technology, which has recently emerged as a major issue in autonomous driving technology. It explores the approach of replacing individual perception, decision-making, and control modules with a single integrated neural network. First, the paper examines the structure and limitations of traditional modular autonomous driving systems, which consist of separate perception, decision-making, and control modules. Then, it analyzes the structure of end-to-end learning autonomous driving systems, highlighting key technologies such as imitation learning and reinforcement learning, training methods, input and output formats, and evaluation methods and metrics. Finally, the paper reviews current technical issues and related structures of end-to-end learning, presenting various research cases and experimental results."
}