面向自动驾驶的感知决策一体化综述  

A review of integrated perception and decision-making for autonomous driving

作　　者：刘泽禹 张慧 李浥东 LIU Zeyu;ZHANG Hui;LI Yidong(School of Computer and Information Technology,Beijing Jiaotong University,Beijing 100044,China)

机构地区：[1]北京交通大学计算机科学与技术学院,北京100044

出　　处：《智能科学与技术学报》2025年第1期4-20,共17页Chinese Journal of Intelligent Science and Technology

基　　金：中央高校基本科研业务费专项资金资助(No.2024JBMC045)。

摘　　要：近年来,感知决策一体化的端到端方案取得了突破性进展,为提升自动驾驶的安全性和可靠性提供了新的思路。现有的综述多聚焦于单独车辆对外部环境的感知,忽略了在多智能体协同下的复杂交互关系。基于当前的研究现状,从应用方法的角度出发,探讨了单智能体感知决策一体化方法,并分析了在多智能体协同条件下的感知决策一体化方法。首先,对感知决策一体化自动驾驶的基础技术进行了总结;其次,重点介绍了感知决策一体化自动驾驶技术的最新研究进展;再次,阐述了自动驾驶领域常用的大规模公共数据集;接着,介绍了目前常用的感知决策一体化自动驾驶评估方法;最后,对感知决策一体化自动驾驶进行了总结与展望。In recent years,end-to-end solutions integrating perception and decision-making have made groundbreaking progress,providing new approaches to enhancing the safety and reliability of autonomous driving.Existing reviews often focus on the perception of the external environment by individual vehicles,overlooking the complex interactions in multi-agent collaboration.Based on the current state of research,integrated perception and decision-making methods for single agents were explored from an application perspective,and methods under multi-agent collaborative conditions were analyzed.Firstly,foundational technologies for integrated perception and decision-making in autonomous driving were summarized.Secondly,the latest research advancements in integrated perception and decision-making technologies for autonomous driving were highlighted.Thirdly,large-scale public datasets commonly used in the field of autonomous driving were elaborated upon.Then,the evaluation methods currently employed for integrated perception and decision-making in autonomous driving were introduced.Finally,a summary and outlook on integrated perception and decision-making in autonomous driving were presented.

关 键 词：智慧交通 自动驾驶 感知决策一体化 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]