摘要
为建立更加简单的换道决策模型和考虑换道车辆和目标车道车辆间的相互作用,在换道效用和安全间隙选择的传统方法基础上,将社会力跟驰模型与换道模型相结合,提出了一种基于社会力的驾驶员主动换道决策行为模型.首先,以社会力模型中跟驰力作为各车道运行效用函数,构建换道目标车道选择效用模型;其次,考虑换道过程车辆纵向安全性,利用跟驰力搭建换道车辆和目标车道车辆间相互作用效用模型以对安全间隙选择进行约束;最后,对所建立的模型利用NGSIM(next generation simulation)数据和MATLAB遗传算法工具箱中genetic algorithm函数对多车道下驾驶员换道决策行为(不换道、向右换道、向左换道)进行标定和验证.研究结果表明:基于社会力的主动换道决策模型能够很好地识别出驾驶员的换道决策行为,最优参数在标定数据中对不换道、向右换道、向左换道的识别率分别达到了93.44%、93.14%和90.77%,验证数据中换道决策行为识别率分别达到了86.16%、80.00%和80.27%;标定和验证的单个识别率都在80.00%以上,整体识别率分别达到92.66%和83.28%.
In order to establish a simpler lane-changing decision-making model and consider the interaction between the lane-changing vehicle and the vehicles of the target lane,on the basis of the traditional methods of lane-changing utility and selection of safety gap,an initiative lane-changing decision-making model for drivers based on the social model by combining the social force following model with the lane-changing model was proposed.First,taking the following force in the social force model as the utility function of each driveway,a lane-changing decision-making utility model for the target lane selection was established.Second,considering the vehicle s longitudinal safety in the lane-changing process,the following force was used to build utility model of interaction between the lane-changing vehicle and the vehicles of the target lane to constrain the selection of safety gap.Finally,the proposed model was calibrated and validated by using next generation simulation(NGSIM)data and genetic algorithm function in MATLAB genetic algorithms toolbox for drivers lane-changing decision-making behaviours(keeping following,changing to right lane,changing to left lane).The study results show that the lane-changing decision-making model based on social force could identify accurately the drivers initiative lane-changing decision-making behaviour.The recognition rates of the optimal parameters in calibration data for drivers lane-changing decision-making behaviours(keeping following,changing to right lane,changing to left lane)reached 93.44%,93.14%,and 90.77%,respectively,and the recognition rates of the lane-changing decision-making behaviours in calibration data reached 86.16%,80.00%,and 80.27%.Each recognition rate in calibration data and validation data was above 80.00%,and the overall recognition rates reached 92.66%and 83.28%,respectively.
作者
杨达
苏刚
吴丹红
熊明强
蒲云
YANG Da;SU Gang;WU Danhong;XIONG Mingqiang;PU Yun(School of Transportation and Logistics,Southwest Jiaotong University,Chengdu 610031,China;Traffic Management Research Institute of the Ministry of Public Security,Wuxi 214151,China)
出处
《西南交通大学学报》
EI
CSCD
北大核心
2018年第4期791-797,共7页
Journal of Southwest Jiaotong University
基金
国家自然科学基金资助项目(51408509
51278429)
中央高校基本科研业务费专项资金资助项目(2682016CX046)
关键词
公路运输
换道决策
社会力
遗传算法
NGSIM数据
highway transportation
lane-changing decision
social force
genetic algorithms
NGSIM data
