摘要
为了应对越来越多的乘客利用手机软件实时查看车辆位置信息以帮助其计划出发时间这一发展趋势,探究实时车辆信息共享程度对公交动态控制策略的影响,结合长短时记忆网络(long short term memory,LSTM)预测模型和遗传算法(genetic algorithm,GA)构建GA-LSTM实时混合公交运行轨迹预测模型;在实时公交运行轨迹预测模型的基础上,建立考虑乘客接受实时车辆信息共享程度、以最小化车头时距偏差和乘客乘车时间为目标的公交动态驻站控制策略;通过西安市实际数据测试实时公交运行轨迹预测模型和公交动态驻站控制策略的有效性,并讨论不同最大驻站控制时间、接受实时信息乘客比例对公交动态驻站控制策略的影响。研究结果表明:利用GA-LSTM预测模型预测停车服务时间和站间运行时间的均方根误差(RMSE)相较于传统LSTM预测模型的预测结果分别降低了21.19%和44.55%;考虑实时车辆信息共享的动态控制策略预测结果比未考虑信息共享的动态控制策略目标函数预测结果降低了31.66%;随着最大驻站时间增加,目标函数值减少到定值后保持稳定;随着接受实时信息乘客比例增加,目标函数值持续减少。研究结果有助于交通管理部门结合乘客接受信息共享程度设计更灵活、有效的公交动态控制策略,从而为改善公交运营状况、提高公交服务水平奠定理论基础。
To respond to the trend that more and more passengers will use real-time location information from mobile phone software to help them plan their departure times,investigate the impact of the degree of real-time vehicle information sharing on the dynamic control strategy,the long short term memory(LSTM)model and genetic algorithm(GA)were used to propose a hybrid GA-LSTM real-time bus trajectory prediction model.A bus control strategy considering information sharing was developed based on the bus trajectory prediction model to minimize the waiting time and onboard travel time.And the performance of the prediction model and strategy was assessed by real-world data in Xi'an.Moreover,the performance of the dynamic control strategy with various maximum holding time and the proportion of passengers receiving real-time information were analyzed.The results show that the prediction result of dwell time and link travel time prediction by the GA-LSTM model outperformed the LSTM model,the average root mean square error(RMSE)value reduce by 21.19%and 44.55%respectively.In addition,the objective value is reduced by 31.66%under real-time information sharing.The value of the objective function decreases to a constant value and then remains stable as the maximum holding time increases,and the value of the objective function continues to decrease as the proportion of passengers receiving real-time information increases.The conclusion can help transport agencies to design flexible bus dynamic control strategies based on the proportion of passengers receiving information sharing,and lay a theoretical foundation for enhancing bus operations and service levels.3 tabs,8 figs,28 refs.
作者
姜瑞森
胡大伟
孙倩
高天洋
JIANG Rui-sen;HU Da-wei;SUN Qian;GAO Tian-yang(School of Transportation Engineering,Chang'an University,Xi'an 710064,Shaanxi,China)
出处
《长安大学学报（自然科学版）》
CAS
CSCD
北大核心
2023年第6期95-105,共11页
Journal of Chang’an University(Natural Science Edition)
基金
陕西省自然科学基础研究计划项目(2021JZ-20)。
关键词
交通工程
动态控制策略
实时预测
GA-LSTM预测模型
信息共享
traffic engineering
dynamic control strategy
real-time prediction
GA-LSTM prediction model
information sharing
