Original title: End-to-end Autonomous Driving using Deep Learning: A Systematic Review
Authors: Apoorv Singh
This article explores the fascinating world of end-to-end autonomous driving, which is a machine learning system that can drive a vehicle by itself. The system takes in raw sensor input data and other relevant information and directly outputs control signals or planned routes for the vehicle. The authors of this article aim to provide a comprehensive review of the various techniques that have been developed to achieve this end-to-end task using machine learning.
The article discusses a range of techniques, such as object detection, understanding the context of scenes, tracking objects, predicting trajectories, planning routes, controlling the vehicle, analyzing social behavior, and establishing communication. Particularly, the authors focus on recent advancements in fully differentiable end-to-end reinforcement learning and deep learning techniques.
The authors also organize the techniques into different categories and present their research trends. Additionally, the article addresses the current challenges in this field and suggests potential future directions for further research. Overall, this article gives readers a detailed understanding of the current state and potential future of end-to-end autonomous driving.
Original article: https://arxiv.org/abs/2311.18636