Deep Learning for Autonomous Driving: A Survey of Methods, Paradigms, and Future Trends

Changsha No.15 High School, Changsha, Hunan Province, China

^* Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract

With the continuous growth of transportation demand and the increasingly prominent issue of road safety, improving traffic efficiency and driving safety has become a core problem that urgently needs to be solved. Although the emergence of autonomous driving technology has brought significant convenience to public transportation, traditional artificial intelligence methods overly rely on manual labeling and fixed models, making it difficult to cope with complex and changing real-world road environments. In scenarios of extreme weather, sudden traffic accidents, or uncertain driving behavior, its performance is particularly inadequate. As the core supporting technology of automatic driving, deep learning has the ability to independently extract key information from massive data, conduct self-training and model iteration, and provides a solid foundation for the intelligence and self-evolution of auto drive system. This paper systematically reviews the typical applications of deep learning in autonomous driving, including environmental perception, behavior prediction, decision-making and control, and analyzes the advantages and limitations of deep learning based-methods such as convolutional neural networks, recurrent neural networks, Transformers. Moreover, the current challenges faced, such as insufficient data, poor model generalization were summarized, and future development directions were discussed.