车辆队列预测巡航控制研究综述

Survey of Predictive Cruise Control for Vehicle Platooning

车辆队列巡航控制主要依据范围有限的交通环境信息,但环境的高度不确定性会影响车辆建模精度与控制效果。预测巡航控制作为巡航控制的一种演进,已成为当前的研究热点。为全面分析车辆队列预测巡航控制的研究进展,从交通环境信息预测、队列运动行为决策、队内车辆轨迹规划、车辆轨迹跟踪控制等4个方面进行概述。首先,介绍车辆队列对交通环境信息的预测研究进展,包括采用车路协同获取前方道路地理、交通等信息,以及通过车载传感器预测周边环境车辆运动状态,重点介绍基于深度学习的轨迹预测方法研究现状及发展趋势;其次,介绍车辆队列协同行为决策问题的研究进展,详细阐述博弈论与机器学习在协同行为决策领域的重要作用,指出模型与数据混合优化的行为决策发展趋势;再次,针对车辆协同轨迹规划问题,从模型驱动与数据驱动2个角度,分别对当前研究进行梳理,并说明强化学习在协同轨迹规划方面具备的优势;然后,从预测巡航控制、车辆跟踪控制等2个方面,分别阐述车辆轨迹跟踪控制问题,并指出基于数据和模型联合驱动的车辆跟踪控制方法具有较大应用潜力;最后,总结车辆队列预测巡航控制的研究现状与不足,并对该领域的未来发展趋势进行展望,为其后续应用提供新思路。

Vehicle platoon cruise control is mainly based on local limited traffic environment information.However,high uncertainty of the environment can affect vehicle modeling accuracy and control performance.As an evolution of cruise control,predictive cruise control has become a current research hotspot.In order to comprehensively analyze the research progress of vehicle platoon predictive cruise control,there are summarized four aspects,i.e.,traffic environment information prediction,platoon motion behavior decision-making,vehicle trajectory planning,and vehicle trajectory tracking control.Firstly,it is introduced the research progress of vehicle platoon prediction of traffic environment information,including geography and traffic information of its road ahead through vehicle-infrastructure cooperation,and predicting motion states of surrounding vehicles through on-board sensors.The status quo and its trends of trajectory prediction methods based on deep learning are mainly introduced;Secondly,it is introduced the progress of decision-making of cooperative vehicle motion behaviors,while emphatically introducing the important roles of game theory and machine learning in this field,and pointing out trends of motion behaviors decision-making using the combined optimization with physical model and data;Thirdly,aiming at the problem of vehicle cooperative trajectory planning,current researches are sorted out from the perspectives of model-driven and data-driven,while advantages of reinforcement learning in collaborative trajectory planning are illustrated;Then,the problem of vehicle trajectory tracking control is expounded from two aspects of predictive cruise control and vehicle tracking control,respectively,while it is pointed out that the vehicle control method jointly driven by model and data has great application potential;Finally,the status quo and shortcomings of vehicle platoon predictive cruise control are summarized,and future trends in this field are prospected to provide new ideas for its application.