Research Article AStudyonTravelTimeEstimationofDivergingTrafficStreamon Highways Based on Timestamp Data Sunghoon Kim , 1 Hwapyeong Yu , 2 and Hwasoo Yeo 2 1 e Korea Transport Institute, Sejong, Republic of Korea 2 Korea Advanced Institute of Science and Technology, Daejeon, Republic of Korea Correspondence should be addressed to Hwasoo Yeo; hwasoo@gmail.com Received 24 March 2020; Revised 24 November 2020; Accepted 19 January 2021; Published 28 January 2021 Academic Editor: Rakesh Mishra Copyright©2021SunghoonKimetal.isisanopenaccessarticledistributedundertheCreativeCommonsAttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Traveltimeisvaluableinformationforbothdriversandtrafficmanagers.Whileproperlyestimatingthetraveltimeofasingleroad section,anissueariseswhenmultipletrafficstreamsexist.Inhighways,thisusuallyoccursattheupstreamofdivergebottleneck. eaimofthispaperistoprovideanewframeworkfortraveltimeestimationofadivergingtrafficstreamusingtimestampdata only.Whileprovidingtheframework,themainfocusofthispaperisonperformingafewanalysesonthestageoftraveltimedata classification in the proposed framework. ree sequential steps with a few statistical approaches are provided in this stage: detectionofdatadivergence,classificationofdivergentdata,andoutlierfiltering.First,adivergencedetectionindex(DDI)ofdata has been developed, and the analysis results show that this new index is useful in finding the threshold of determining data divergence. Second, three different methods are tested in terms of properly classifying the divergent data. It is found that our modified method based on the approach used by Korea Expressway Corporation shows superior performance. ird, a polynomialregression-basedmethodisusedforoutlierfiltering,andthisshowsreasonableperformanceevenatarelativelylow marketpenetrationrate(MPR)ofprobevehicles.en,theoverallperformanceofthetraveltimeestimationframeworkistested, and this test demonstrates that the proposed framework can show improved performance in distinctively estimating the travel times of two different traffic streams in the same road section. 1. Introduction e growth in urban population and city-centred life pat- ternshasraisedaseriesofproblemsinmoderncitiessuchas trafficcongestionandaccidents.Totackletheseissues,there have been numerous attempts to implement intelligent transportationsystem(ITS)intheroadnetworks.Inthefield of ITS, travel time is one of the most valuable information for both vehicle drivers and traffic managers. In advanced travelerinformationsystem(ATIS),updatingtraveltimefor driversinrealtimeenablesthemtomakeinformeddecisions on their route choices to avoid congested roads [1]. In advanced traffic management system (ATMS), based on proper analyses of traffic states in relation to travel time information, traffic managers can develop various control and operational strategies to reduce road congestion [2]. Furthermore, travel time information can be used as supplementary input for further development of the recent studies related to traffic flow analysis [3], including short- term prediction using neural network techniques [4, 5] and long-termforecastsusingfuzzytheory[6,7].Hence,proper estimationoftraveltimeoneachroadinareal-timemanner is crucial for further development and implementation of ITS. Travel time information can be obtained indirectly or directly.eindirectwaysusuallyestimatetraveltimebased on traffic flow and speed measured by point sensors on the roadside such as loop detectors, video cameras, and radars [8–10]. e direct ways measure the travel time using rec- ordsattollgatesandroadsideunits(RSU),whicharepartsof automaticvehicleidentification(AVI)technologies[11].e useofGlobalPositionalSystem(GPS)ontheprobevehicles equippedwithnavigationdevicesorsmartphonesisanother directwayoftraveltimemeasurement,andthisistherecent Hindawi Journal of Advanced Transportation Volume 2021, Article ID 8846634, 13 pages https://doi.org/10.1155/2021/8846634