摘要
从控制架构与驾驶权决策出发,阐述了人机共驾的研究现状以及发展趋势;在控制架构方面,分析了切换控制架构和共享控制架构的特点和应用范围,并提出了混杂控制架构概念;在驾驶权决策方面,讨论了不同驾驶权决策方法对不同来源、不同性质信息的使用方式,概括了执行驾驶权分配时直接和间接共享控制方式所涉及的方法,梳理了策略层决策与执行层决策的研究角度与方法。研究结果表明:针对高阶自动驾驶上路运行安全问题,发展混杂控制架构对安全员干预场景下的系统动态进行描述有利于避免模型失配,从而为控制性能优化和稳定性设计提供了基础;通过融合全息态势感知与数据智能的方式收集和整合多个信息源的数据,能够更加全面地理解人机共驾系统中诸多要素的动态变化并做出最优驾驶权决策;相较于直接共享控制,间接共享控制能避免人机控制流直接对抗,但是其动态驾驶权分配执行层面不仅需要考虑人机之间的冲突反馈,还需要确保合理的交互体验以体现间接共享控制的优势;基于智能体的策略层决策方法并不依赖于数学模型精度,能够自适应环境的动态变化;基于博弈论的执行层决策方法通过建模人机交互过程能够增强驾驶权决策系统的可控性和可解释性;未来的人机共驾系统设计应进一步优化交互体验,关注发展平等共融的人机关系,并提高控制系统的鲁棒性以及驾驶权决策的可解释性和适应性。
In view of the control architecture and driving authority decision-making,the research status and development trend of driver-automation cooperative driving were expounded.In terms of control architecture,the characteristics and application range of switching control architecture and shared control architecture were analyzed,and the concept of hybrid control architecture was proposed.In terms of driving authority decision-making,the ways of using different sources and natures of information in different driving authority decision-making methods were discussed.The methods involved in the direct and indirect shared control methods when implementing the allocation of driving authority were summarized.The research perspectives and methods of decision-making at the strategy level and the executive level were sorted out.Research results show that for the safety problems of high-level automated driving on the road,the development of hybrid control architecture for describing the system dynamics under human safety intervention scenarios is conducive to avoiding model mismatch,which provides the foundation for control performance optimization and stability design.By integrating holographic situational awareness and data intelligence to collect and integrate data from multiple information sources,the dynamic changes of many factors in the driver-automation cooperative driving system can be more comprehensively understood,and the optimal driving authority decision can be made.Compared with direct shared control,indirect shared control can avoid direct confrontation between driver and automation control flows.However,at the executive level of dynamic driving authority allocation,it is necessary to consider the conflict feedback between driver and automation and ensure a reasonable interactive experience,so as to reflect the advantages of indirect shared control.The decision-making method based on the agent at the strategy level is independent of the accuracy of the mathematical model and can adapt to the dynamic change of the environment.The decision-making method based on game theory at the executive level can enhance the controllability and explainability of the driving authority decision-making system by modeling the driver-automation interaction process.In the future,the driver-automation cooperative driving system should be designed to further optimize the interactive experience.Meanwhile,the development of equal and inclusive driver-automation relationships is necessary.The robustness of the control system and the interpretation and adaptability of driving authority decision-making should be improved as well.
作者
黄炜
黄起鹏
HUANG Wei;HUANG Qi-peng(School of Mechanical Engineering and Automation,Fuzhou University,Fuzhou 350108,Fujian,China)
出处
《交通运输工程学报》
北大核心
2025年第1期48-65,共18页
Journal of Traffic and Transportation Engineering
基金
福建省自然科学基金项目(2021J01559)
国家自然科学基金项目(52272389)。
关键词
人机共驾
混杂控制
驾驶权
交互协同
强化学习
博弈论
driver-automation cooperative driving
hybrid control
driving authority
interactive collaboration
reinforcement learning
game theory
