Big Data in Capturing Travel Time

Transcription

1 Big Data in Capturing Travel Time A quick snapsht f the applicatins in Auckland Transprt Authr/Presenter: Bill Qu, B.E.(hns), M.E.(hns) Principal Traffic Engineer, Auckland Transprt bill.qu@aucklandtransprt.gvt.nz C-Authr: Miguel Menezes, B.Sc. (Civil Engineering) Team Leader Operatinal Planning and Perfrmance, Auckland Transprt Miguel.Menezes@aucklandtransprt.gvt.nz

2 ABSTRACT Time is mney! The ability t quantify jurney time reliability in terms f travel time is becming increasingly pssible in the mdern age. Fr transprtatin practitiners, this t is becming valuable infrmatin that enables better ptimisatin f the transprt netwrk t be made. What nt s lng ag started as flating vehicle surveys, has quickly becme BIG DATA. S hw d we capture the data? Hw we can use this data? What can we learn frm this? And hw helpful is it really? What new thinking is being applied t jurney infrmatin gathering - hw we are making the mst f the technlgical advances in this field t infrm planners, engineers, decisin makers and ultimately ur custmers. This practical paper will prvide examples frm the Auckland cntext and the variety f uses frm netwrk perfrmance mnitring, netwrk peratins planning, and temprary traffic management (including majr new prjects). Hw we are telling the true stry in amidst the squeaky wheels!

3 1.0 Intrductin Understanding the traffic perfrmance n ur netwrk has always been an integral part f respnsivities fr transprt practitiners. The infrmatin can be used in varius areas, including: Strategy and Transprt planning Operating Perfrmance Reprting Operatinal Deficiency Analysis BCR assessment fr prject pst-implementatin Prviding custmers with better and mre accurate infrmatin Technlgy is rapidly grwing, s are ur ways f capturing the travel time infrmatin. What was cnsidered the nly reliable way a few years back may deemed redundant nwadays. This paper aims t prvide a snapsht f ppular technlgies and methdlgies available currently t capture the travel time infrmatin. It then described a few case studies and applicatins in Auckland Transprt, in terms f hw we are making use f the data. The BIG DATA cncept we referred in this paper is nt the traditinal IT terminlgy. In this reprt, the BIG DATA is t truly reflect the huge amunt f data available t us as transprt practitiners. It prbably is the glden age fr traffic data mining, and it may keep grwing fr the next many years. 2.0 Capturing the Infrmatin Traditinally, back five t seven years ag, flating vehicle surveys were still the predminant methd f capturing travel time infrmatin. This methd invlves having dedicated drivers travelling alng selected rutes and crridrs n the netwrk, back and frth, t mimic a snapsht f the typical travel times. It is cnsidered highly accurate fr the time perid and rutes cvered, in fact until nw, many times the transprtatin practitiners have been using the result f flating vehicle surveys t calibrate ther different technlgies. Hwever, t have a wider and better understanding f travel perfrmance frm a netwrk perspective, a few inevitable drawbacks were present: Data sample size is small n selected rutes nly, especially if the rutes were t lng, r nt t many drivers were deplyed fr the survey typically between five and ten drivers were used. The survey is limited t a daily (r a very few days) snapsht. It may nt be the typical perfrmance f the crridrs, and wuld nt be able t track mnthly and seasnal variatin. Inflexibility during accident r unexpected event. Once survey is cmmitted, there is n way ging back. If the weather cnditins changes, r accident happens n the

4 netwrk, the survey results wuld nt be reflecting a typical day, and hence nrmally cnsidered nt accurate. Nt value fr mney, r even very cstly. The amunt f manual wrk invlved, frm planning and rganising the survey, checking the cnditin n the survey day (weather and accident etc.), prcessing the data, all rely heavily n peple, hence nrmally manual surveys have high labur cst. Nwadays, with the rapid advancement f technlgy, different applicatins and methds were develped t better capture the travel time infrmatin. In general, technlgy can be divided int tw ways: Matching r Tracking: Matching typically invlves at predefined lcatins (nrmally majr intersectins), certain unique identificatins, either frm the vehicles r devices in the vehicle, were captured and matched, technlgies includes: ANPR (Autmatic Number Place Recgnitin): matching the number place f vehicles. Typically the kerbside lane is used fr easy setup the camera. Bluetth r Wi-Fi: matching the MAC (media access cntrl) address frm any cmmunicatin device in the car. Nrmally the Matching technlgy requires permanent r temprary infrastructure installatin. It is cnsidered that higher data sample size is bserved using the Matching technlgy The disadvantage is the transprtatin practitiners culd nt be able t tell what happened between the tw matching pints, including start/stps, r the actual rutes taken t arrive between the tw marching pints (rat running) hwever this may be less significant if the data sample size is large enugh. Tracking technlgy means cnstantly tracking the vehicle r devices in the vehicles. The frequency f tracking varius depends n the technlgy, and the individual device settings: GPS tracking: everything can be tracked these days peple, pets, yur mbile phne. And vehicle is n exceptin. Originally used fr fleet management, this technlgy has been deplyed widely and cmmnly t assist capturing travel time infrmatin. If apprpriate GPS tracking devices are installed in the vehicles, the devices will reprt back cnstantly (in secnds) the vehicle s lcatin and driving data back t the satellite, smetimes in real time. Mbile phne data tracking remember the times when yur app. n yur smartphne ask yu t share yur lcatin data? Well, that infrmatin may have cntributed t generate the cngestin infrmatin. Very similar t GPS tracking, the mbile phnes are cnstantly sharing the lcatin data back t the satellite. Tracking technlgy culd hpefully give transprtatin practitiners mre cnfident abut data accuracy as they can apply filtering algrithm t take ut utliers. Hwever, tw ptential prblems f using tracking technlgy: In general, relatively lw data sample size cmparing t matching technlgy, especially after filtering

5 Canyn effect with high rise surrundings. The signal sent t the satellites may be bunced arund, reflecting lw accurate lcatin infrmatin. This als applies in tunnels. There are ther technlgies exist in the market, including algrithms t use lps and/r radars. These are well tested and utilised alng the mtrway netwrk, hwever the true applicatin n the arterial netwrk is still unclear, hence this paper will nt discuss these in detail. The table belw summarised the abve paragraphs: Type Technlgy Ntes Traditinal Flating Vehicle Surveys Deemed highly accurately Small data sample size Typicality highly relies n weather, netwrk cnditin during the survey perid Can t track mnthly/seasnal variatin High peratinal cst Mre used as a validatin methd Matching ANPR (number plate) Bluetth r Wi-Fi (MAC address) Typically, larger data sample size than Tracking Nrmally requires permanent r temprary infrastructure installatin. Upfrnt cst fr setup and n-ging maintenance. Nt be able t tell what happened in between matching pints hwever less significant if data sample size is large enugh Tracking GPS Mbile phne data Others Lps//radar + calculatin algrithms Culd apply filtering algrithm t easily take ut utliers, data accuracy higher Nrmally purchase services frm third party. May struggle with data sample size Canyn effect in high rise surrundings, als tunnels Nt discussed in this paper

6 3.0 Making use f the infrmatin Case Studies First and fremst, it is imprtant t nte that except flating vehicle survey, we are bligated t receive and prcess the aggregated data nly all the technlgy and algrithms applied need t make sure they satisfy the Privacy Act After infrmatin gathering, filtering, prcessing and analysing, it can then be used in many areas, including netwrk perfrmance mnitring, netwrk peratins planning, and temprary traffic management (including majr rad wrks) mnitring. In this chapter, we will be presenting a few case studies f hw Auckland Transprt has been making use f the data frm multiple surces, fr varius prjects. 3.1 Netwrk Perfrmance Mnitring Auckland Transprt has been leading n reprting netwrk perfrmance using technlgy we started six years ag using GPS tracking. By wrking clsely with a lcal fleet management cmpany, tgether we have develped a rbust algrithm, prcessed infrmatin validated using flating vehicle surveys, t prvide regular netwrk perfrmance snapsht, as well as ad-hc requests fr detailed studies and prjects. We have managed t cver all f the primary arterial netwrks, as well as mst f the secndary arterials including cllectrs. The netwrk cverage culd als be expanded if required n an ad-hc basis. By prcessing the travel time infrmatin, we have prduced the cngestin maps, as well Auckland Transprt defined KPIs relating t the netwrk perfrmance mnitring:

7 Thrugh prcessing the aggregated data, Auckland Transprt has develped a rbust and sphisticated system, t reprt n the pre-defined Key Perfrmance Indicatrs (KPIs), including: Average netwrk speed Htspt (similar t the blackspt cncept in rad safety) Level f Service (LOS) table, hrizntal cmparisns acrss different mnths Traffic Delays Jurney Reliability Similarly, we have prduced similar netwrk perfrmance screcard fr freight, buses and pedestrians:

8

9 3.2 Netwrk Operatins Planning With all the perfrmance KPIs defined and captured regularly fr varius mdes, we have been able t greatly cntribute twards varius prjects, including: Strategy and Planning Crridr Management Plan (e.g. Hillsbrugh Rad CMP) Rutine Traffic Signal Optimisatin (e.g. New Nrth Rad Optimisatin) Multi-mdal Deficiency Assessment fr buses and vehicles (e.g. Parnell Rad Bus Lane Study) Better understanding and hence respnding t the custmers fr cngestin related enquiries Perfrmance tracking and reprting pst prject implementatin (e.g. Fanshawe Street Bus Lane) Data feed t Netwrk Operating Plans and SmartRads sftware fr GAP assessment Crridr Management Plans The multi-mdal netwrk perfrmance mnitring infrmatin was a key part f the crridr management plan prjects. It identifies the key multi mdal deficiencies frm an peratinal level. Belw is a snapsht frm the Hillsbrugh Crridr Management Plan.

10 3.2.2 Traffic Signal Rutine Optimisatin We have started t intrduce and include the multi-mdal cngestin infrmatin int the traffic signal rutine ptimisatin. T assist the ATOC signal engineers where the ptential htspts are n the rutes being ptimised. Belw is a screensht f example fr New Nrth Rad traffic signal rutine ptimisatin Multi-mdal Deficiency Assessment fr buses and vehicles Thrugh ur business as usual netwrk perfrmance reprting, we are able t identify prjects thrugh deficiency assessment. Parnell Rad bus lane is an example. The mnthly bus cngestin mnitring indicates severe cngestin alng Parnell Rad suthbund in the afternn peak. Thrugh ur investigatin, we have intrduced a quick win prjects t cnvert the kerbside parking int an afternn peak bus lane. The imprvement is easily visible (the darker the clur, the wrse the LOS, hence the mre severe cngestin is).

11 Data feed t Netwrk Operatins Planning (NOP) tl (SmartRads) The multi-mdal netwrk perfrmance infrmatin can be directly fed int the Netwrk Operatins Planning tl (SmartRads sftware) as well. It has been well used fr varius prjects planning including traffic signal rutine ptimisatin. Belw is an example f NOP prduced fr Greenlane crridr traffic signal ptimisatin, based n the peratinal multimdal perfrmance we have been reprting.

12 Temprary Perfrmance Tracking n Majr Prjects - CRL Travel time tracking has als been applied t majr prjects, the City Rail Link is ne f the examples. Apart frm the high level Ntice Of Requirement cnditins set up fr the resurce cnsent applicatin, Auckland Transprt has been wrking with AraFlw, t setup Bluetth travel time tracking n the key crridrs in the CBD. Alarm systems were als setup when the travel time experienced went ver certain pre-defined benchmarks. This has given Auckland Transprt the pprtunity t access near real time travel time infrmatin, actin quickly if needed t deal with the (unexpected) traffic cnditins. Since Octber 2015, Auckland Transprt has been reprting n the netwrk perfrmance snapsht, nrmally n a mnthly basis, but als n ad-hc checks when necessary.

13 Cnclusin With the fast explsin f technlgy, there are definitely mre, and ptentially better ways f capturing travel time infrmatin than the traditinal flating vehicle surveys. With this rapid develpment, we need t be aware f the prs. and cns. f varius technlgies, and chse the mst apprpriate ne accrding t ur requirement. Auckland Transprt has been testing and adpting new technlgy since five t six years ag. We have prcured, validated varius technlgies, and applied them fr different uses. There is a grwing demands fr accurate data, it is imprtant that all transprtatin practitiners share their relevant experience. There is n right r wrng technlgies, it all depends n hw we use it. Technlgy wise, we culd never really be able t predict what wuld happen in the next few years. Face recgnitin may becme ppular, r we will all be planted with chips t ur bdies. Wh knws. The best we can d is t understand the new technlgy, adapt t it, make gd use f it, and make sense ut it.