Due to the speed difference and the complex interaction between merging and throughlane vehicles at freeway merging sections,crashes involving both human drivers and automated vehicles(AVs)persist.To assist AVs to pre...Due to the speed difference and the complex interaction between merging and throughlane vehicles at freeway merging sections,crashes involving both human drivers and automated vehicles(AVs)persist.To assist AVs to predict the intentions of surrounding vehicles for further dynamic motion planning,researchers have focused on developing trajectory prediction algorithms.Few studies,however,have developed merging trajectory prediction models using naturalistic driving data in China,making it urgent to put it on the agenda for AVs’safety and efficiency at freeway merging sections.Based on the merging periods extracted from the Shanghai Naturalistic Driving Study(SH-NDS),this study compares merging behavior on freeways with through-lane speed limits of 80 km/h,100 km/h,and 120 km/h using analysis of variance(ANOVA).Merging trajectory prediction algorithms for these three speed limit cases are trained and tested using backpropagation neural network(BPNN)and long short-term memory neural network(LSTMNN)approaches.Results show that:1)there are significant differences among the three cases in all merging behavior variables except for longitudinal gap,and 2)the BPNN algorithm for merging trajectory prediction demonstrates superior performance compared to the LSTMNN.Two major contributions to the safe operation of AVs are provided:1)the developed algorithms can be integrated into AV systems to accurately predict real-time desired trajectories of nearby merging vehicles in uncongested traffic conditions,and assist ongoing motion planning strategies for AVs;2)the algorithms can be incorporated in simulation tests for AV safety evaluation involving freeway merging sections.展开更多
基金sponsored by the National Natural Science Foundation of China(No.51878498)the Belt and Road Cooperation Program under the 2023 Shanghai Action Plan for Science,Technology and Innovation(No.23210750500).
文摘Due to the speed difference and the complex interaction between merging and throughlane vehicles at freeway merging sections,crashes involving both human drivers and automated vehicles(AVs)persist.To assist AVs to predict the intentions of surrounding vehicles for further dynamic motion planning,researchers have focused on developing trajectory prediction algorithms.Few studies,however,have developed merging trajectory prediction models using naturalistic driving data in China,making it urgent to put it on the agenda for AVs’safety and efficiency at freeway merging sections.Based on the merging periods extracted from the Shanghai Naturalistic Driving Study(SH-NDS),this study compares merging behavior on freeways with through-lane speed limits of 80 km/h,100 km/h,and 120 km/h using analysis of variance(ANOVA).Merging trajectory prediction algorithms for these three speed limit cases are trained and tested using backpropagation neural network(BPNN)and long short-term memory neural network(LSTMNN)approaches.Results show that:1)there are significant differences among the three cases in all merging behavior variables except for longitudinal gap,and 2)the BPNN algorithm for merging trajectory prediction demonstrates superior performance compared to the LSTMNN.Two major contributions to the safe operation of AVs are provided:1)the developed algorithms can be integrated into AV systems to accurately predict real-time desired trajectories of nearby merging vehicles in uncongested traffic conditions,and assist ongoing motion planning strategies for AVs;2)the algorithms can be incorporated in simulation tests for AV safety evaluation involving freeway merging sections.
