A survey of deep learning techniques for autonomous driving

By Sorin Mihai Grigorescu, Bogdan Trasnea, Tiberiu T. Cocias, Gigel Macesanu. 2019

1,743 citations recorded by the source corpus.

The last decade witnessed increasingly rapid progress in self‐driving vehicle technology, mainly backed up by advances in the area of deep learning and artificial intelligence (AI). The objective of this paper is to survey the current state‐of‐the‐art on deep learning technologies used in autonomous driving. We start by presenting AI‐based self‐driving architectures, convolutional and recurrent neural networks, as well as the deep reinforcement learning paradigm. These methodologies form a base for the surveyed driving scene perception, path planning, behavior arbitration, and motion control algorithms. We investigate both the modular perception‐planning‐action pipeline, where each module is built using deep learning methods, as well as End2End systems, which directly map sensory information to steering commands. Additionally, we tackle current challenges encountered in designing AI architectures for autonomous driving, such as their safety, training data sources, and computational hardware. The comparison presented in this survey helps gain insight into the strengths and limitations of deep learning and AI approaches for autonomous driving and assist with design choices.

Subfields: Automotive Engineering, Computer Vision and Pattern Recognition, Social Psychology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Automotive Engineering

In Computer Vision and Pattern Recognition

In Social Psychology

Citations in the corpus

Learning to Prompt for Continual Learning (2021, methodology): “Moreover, the method could be deployed in safety-critical applications, such as autonomous driving systems [15], which may present potential security issues in terms of adversarial attacks [33].”
Driving with LLMs: Fusing Object-Level Vector Modality for Explainable Autonomous Driving (2023, background): “Modern scalable autonomous driving systems, whether they adopt an end-to-end approach using a single network [4], or a component-based configuration that combines learnable perception and motion planning modules [5] [6], face common challenges.”
Advances in adversarial attacks and defenses in computer vision: A survey (2021, background): “With highly active deep learning-based vision research for autonomous vehicles [17], face recognition [18], [19], robotics [20] and surveillance systems [21] etc.”
A Comprehensive Survey of AI-Generated Content (AIGC): A History of Generative AI from GAN to ChatGPT (2023, background): “On the contrary, AIGC for high-stakes applications, including healthcare [248], financial services [249], autonomous vehicles [250], and science discovery [251], are still challenging.”
Vision-Language Models for Vision Tasks: A Survey (2023, background): “, image classification, object detection and semantic segmentation) is a long-standing challenge in computer vision research, and it is also the cornerstone of a myriad of computer vision applications in autonomous driving [1], [2], remote sensing [3], [4], robotics [5], [6], etc.”
AdaBits: Neural Network Quantization With Adaptive Bit-Widths (2019, background): “For example, popular mobile apps such as TikTok and Snapchat on portable devices pose stringent requirements on response latency and energy consumption, while visual recognition system embedded in a self-driving vehicle [13, 40] is more demanding on fast and accurate prediction.”
Milestones in Autonomous Driving and Intelligent Vehicles: Survey of Surveys (2023, neutral): “All the surveys are categorized into several sub-sections, including the overall [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], localization [14], [15], [16], [17], [18], [19],”
Explainable Artificial Intelligence: Objectives, Stakeholders, and Future Research Opportunities (2020, background): “…holds true not only in medical diagnostics, but also in processes of job recruitment (Dastin, 2018), credit scoring (Wang et al., 2019), prediction of recidivism in drug courts (Zolbanin et al., 2019), or as autopilots in aviation (Garlick, 2017) and autonomous driving (Grigorescu et al., 2020).”
Machine learning and deep learning (2021, background): “object counting with fields of application ranging from crop detection for yield prediction to autonomous driving (Grigorescu et al. 2020).”