Contents lists available at ScienceDirect Transportation Research Part C journal homepage: www.elsevier.com/locate/trc Dynamic shared autonomous taxi system considering on-time arrival reliability Zhiguang Liu a, ⁎ , Tomio Miwa b , Weiliang Zeng c , Michael G.H. Bell d , Takayuki Morikawa e a Department of Civil Engineering, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan b Institute of Materials and Systems for Sustainability, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan c School of Automation, Guangdong University of Technology, Guangzhou, Guangdong 510006, China d Institute of Transport and Logistics Studies, Business School, The University of Sydney, Sydney, NSW, Australia e Institute of Innovation for Future Society, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan ARTICLEINFO Keywords: Shared autonomous taxi system On-time arrival reliability Historical travel time information System-beneficial path Travel time ABSTRACT Dynamicsharedautonomoustaxi(SAT)systemsareregardedasapromisingmeansofimproving travel flexibility. With no human drivers, SATs urgently require precise traffic information in order to plan accurate paths independently; in addition, on-time arrival is an essential service quality in SAT systems. In this study, taxis are assumed to be replaced with ride-sharing au- tonomous vehicles. To improve the probability of on-time arrival, the reliable path concept and collected travel time information are used to facilitate path finding for SATs, and the potential benefitsareexamined.Twosimulationscenarios—onebasedonhistoricaltrafficinformationand the other based on real-time traffic information—are executed to evaluate the information’s usefulnessinreliablepathfinding.Insimulationresults,reliablepathscenariosshowedahigher on-timearrivalratiothanshortestpathscenarios,inwhichtheshortestpathalgorithmisusedin path finding for SATs, and the historical information-based scenarios showed a higher on-time arrival ratio than the real-time information-based scenarios. A system-beneficial path finding method is proposed and is verified to be effective for mitigating road network congestion. 1. Introduction Autonomousvehicletechnologyhasexperiencedrapiddevelopmentinrecentyears(Caoetal.,2017).TheSocietyofAutomobile Engineers (SAE) International defines five levels of autonomous driving, from driver assistance to full automation (SAE, 2016). To date,manyvehicleshavefulfilledlevel1(driverassistance)and2(partialautomation),andseveralautomobilemanufacturesandIT companies worldwide are now implementing level 4 (high automation) tests (Litman, 2018). A Singaporean technology company, nuTonomy,usesautonomousvehiclesastaxiswithinanareaof2.5squaremiles(Patel,2016).Autonomousvehiclesareexpectedto provide independent mobility for non-drivers, including disabled people and adolescents, while reducing transportation costs, en- vironmental impacts, and congestion (Bansal et al., 2016). Autonomous vehicles can also support ridesharing services such as dy- namicsharedautonomoustaxi(SAT)systems,inwhichautonomousvehiclesareusedassharedtaxis.BecauseSATscanservemore thanonecustomer,theycanincreasetravelflexibility.CustomersrequestSATsviasmartphones,andSATsareassignedtocustomers https://doi.org/10.1016/j.trc.2019.04.018 Received 18 May 2018; Received in revised form 17 April 2019; Accepted 17 April 2019 ⁎ Corresponding author. E-mail addresses: liu.zhiguang@i.mbox.nagoya-u.ac.jp (Z. Liu), miwa@nagoya-u.jp (T. Miwa), weiliangzeng@gdut.edu.cn (W. Zeng), michael.bell@sydney.edu.au (M.G.H. Bell), morikawa@nagoya-u.jp (T. Morikawa). Transportation Research Part C 103 (2019) 281–297 0968-090X/ © 2019 Elsevier Ltd. All rights reserved. T