Hybrid operations of human driving vehicles and automated vehicles with data-driven agent-based simulation. (September 2020)

• Record Type: Journal Article
• Authors: Yao, Fugen; Zhu, Jiangtao; Yu, Jingru; Chen, Chuqiao; Chen, Xiqun (Michael)
• Is Part Of: Transportation research. Volume 86(2020)
• Publication Date: 2020-09
• Subjects: Hybrid ride-hailing market -- Automated vehicle -- Data-driven decision model -- Agent-based modeling and simulation -- Transportation environment; Transportation -- Research -- Periodicals; Transportation -- Environmental aspects -- Periodicals; 354.76
• DOI: 10.1016/j.trd.2020.102469
• Languages: English
• ISSNs: 1361-9209

Abstract

Highlights: Propose data-driven agent-based modeling and simulation (D 2 ABMS) for large-scale transportation networks. Employ data-driven multi-objective deep learning to learn ride-sourcing drivers' offline/online behavior. Collect ride-sourcing data to train and validate the drivers' decision-making model. Test hybrid operations of human driving vehicles and automated vehicles in various scenarios. Explore the environmental impacts of automated vehicles in hybrid ride-hailing market. Abstract: Automated vehicles (AVs) receive tremendous attention and achieve rapid development. It is foreseeable that hybrid operations of human driving vehicles and automated vehicles in the urban transportation environment will be a long-standing state. To investigate the influence of AVs on the hybrid ride-hailing market, data-driven agent-based modeling and simulation (D 2 ABMS) for large-scale transportation networks is proposed, in which human drivers, automated vehicles, and passengers form three types of agents. D 2 ABMS goes beyond existing approaches by employing data-driven multi-objective deep learning to learn ride-sourcing drivers' offline/online behavior. E mbedding is used to represent the hidden attributes of different classes of drivers. Ride-sourcing data collected from the city of Hangzhou, China, are used to train and validate the drivers' decision-making model. Hybrid operations of human driving vehicles and automated vehicles with D 2 ABMS are comprehensively tested in various scenarios. The results show that a small proportion of automated vehicles in the hybrid ride-hailing market can significantly reduce the average waiting time of passengers. Besides, compared to the human driving scenario, the total exhaust emissions and vehicle kilometers traveled can be reduced by 12.3% in the AVs scenario. The proposed D 2 ABMS system has the potential to help transportation planners and ride-hailing platforms to assess their policies and operations management strategies in the era of shared mobility and automated vehicles.