[论文解读] Vehicle-to-Everything Cooperative Perception for Autonomous Driving

Tao Huang, Jianan Liu+6 more|arXiv (Cornell University)|Oct 5, 2023

Advanced Neural Network Applications被引用 20

一句话总结

本论文综述了通过V2X通信实现的自动驾驶协作感知（CP）的演进，提出一个通用的CP框架，并讨论使能技术、数据集、仿真器、挑战与未来方向。

ABSTRACT

Achieving fully autonomous driving with enhanced safety and efficiency relies on vehicle-to-everything cooperative perception, which enables vehicles to share perception data, thereby enhancing situational awareness and overcoming the limitations of the sensing ability of individual vehicles. Vehicle-to-everything cooperative perception plays a crucial role in extending the perception range, increasing detection accuracy, and supporting more robust decision-making and control in complex environments. This paper provides a comprehensive survey of recent developments in vehicle-to-everything cooperative perception, introducing mathematical models that characterize the perception process under different collaboration strategies. Key techniques for enabling reliable perception sharing, such as agent selection, data alignment, and feature fusion, are examined in detail. In addition, major challenges are discussed, including differences in agents and models, uncertainty in perception outputs, and the impact of communication constraints such as transmission delay and data loss. The paper concludes by outlining promising research directions, including privacy-preserving artificial intelligence methods, collaborative intelligence, and integrated sensing frameworks to support future advancements in vehicle-to-everything cooperative perception.

研究动机与目标

• Trace the evolution of cooperative perception (CP) from early ideas to modern V2X-enabled approaches.
• Introduce a generic framework for V2X CP to clarify system components and workflows.
• Categorize CP methodologies by the practical issues they address and assess datasets and simulators.
• Discuss four enabling V2X technologies (edge computing, blockchain, digital twins, 6G) and outline open challenges and future directions.

提出的方法

• Review historical and technical developments in CP and V2X communications from 2008 to mid-2023.
• Present a unified, generic CP framework illustrating multi-agent collaboration and data fusion workflows.
• Propose a taxonomy for V2X CP solutions focusing on practical driving scenarios and system modules.
• Summarize existing datasets and simulators used for CP research and evaluate their roles.
• Discuss open challenges in perception, communication, privacy, security, and policy implications, and propose future research directions.

实验结果

研究问题

• RQ1 How has CP evolved with advancements in V2X communications and AI/learning-based perception?
• RQ2 What is a unified framework that captures CP workflows and system components for V2X-enabled autonomous driving?
• RQ3 What are the key enabling V2X technologies shaping CP (edge, blockchain, digital twins, 6G) and their roles and challenges?
• RQ4 What datasets and simulators are prevalent for CP research, and what gaps remain?
• RQ5 What open challenges and directions are critical for practical deployment of V2X CP in CDA?

主要发现

• CP with V2X substantially addresses occlusion, limited FoV, and sensor reliability limitations of single-vehicle perception.
• A new taxonomy and a generic CP workflow framework are proposed to organize CP research.
• Four enabling V2X technologies—edge computing, blockchain, digital twins, and 6G—are identified as pivotal to CP viability.
• Hybrid and intermediate collaboration schemes are discussed as trade-offs between perception accuracy and communication bandwidth.
• Open challenges across perception, communication, data trust, privacy, and policy are highlighted with guidance for future research.