shared autonomous taxi system and utilization of collected...

Research ArticleShared Autonomous Taxi System and Utilization ofCollected Travel-Time Information

Zhiguang Liu 1 Tomio Miwa 2 Weiliang Zeng 3

Michael G H Bell4 and Takayuki Morikawa5

1Department of Civil Engineering Nagoya University Nagoya 464-8603 Japan2Institute of Materials and Systems for Sustainability Nagoya University Nagoya 464-8603 Japan3School of Automation Guangdong University of Technology Guangzhou Guangdong 510006 China4Institute of Transport and Logistics Studies Business School The University of Sydney Sydney NSW Australia5Institute of Innovation for Future Society Nagoya University Nagoya 464-8603 Japan

Correspondence should be addressed to Weiliang Zeng weiliangzenggduteducn

Received 29 December 2017 Revised 19 June 2018 Accepted 31 July 2018 Published 8 August 2018

Academic Editor Vinayak Dixit

Copyright copy 2018 Zhiguang Liu et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Shared autonomous taxi systems (SATS) are being regarded as a promising means of improving travel flexibility Each sharedautonomous taxi (SAT) requires very precise traffic information to independently and accurately select its route In this studytaxis were replaced with ride-sharing autonomous vehicles and the potential benefits of utilizing collected travel-time informationfor path finding in the new taxi system examined Specifically four categories of available SATs for every taxi request wereconsidered currently empty expected-empty currently sharable and expected-sharable Two simulation scenariosmdashone based onhistorical traffic information and the other based on real-time traffic informationmdashwere developed to examine the performance ofinformation use in a SATS Interestingly in the historical traffic information-based scenario the mean travel time for taxi requestsand private vehicle users decreased significantly in the first several simulation days and then remained stable as the number ofsimulation days increased Conversely in the real-time information-based scenario the mean travel time was constant As the SATfleet size increased the total travel time for taxi requests significantly decreased and convergence occurred earlier in the historicalinformation-based scenario The results demonstrate that historical traffic information is better than real-time traffic informationfor path finding in SATS

1 Introduction

Autonomous vehicles (AVs) have undergone rapid develop-ment in recent years Many automobile manufacturers andIT companies around the world including Google UberTesla and Toyota are testing their AV products on realroad networks [1] In some countries AVs are allowed toenter select areas but not the entire public road network ASingaporean technology company called ldquonuTonomyrdquo usesAVs as taxis within an area of 25 square miles [2] OnDecember 2 2017 four self-driving buses were tested onpublic roads in Shenzhen China which is believed to be thefirst live test of autonomous buses in the world [3] Thereis considerable evidence to indicate that the use of AVs willbecome more widespread in the near future

AVs can be designed to connect with each other and canexchange traffic information related to the road network [4]The AVs start when orders are transmitted to the controllerinside the vehicles and they drive automatically based on theinstructions from the controller Owing to these advantagesAVs are expected to make transportation more efficient andcomfortable and reduce cost environmental impact andcongestion [5]

In a traditional taxi operation system a driver can serveonly a single customer or a group of customers in a point-to-point journey For example the average occupancy rate of ataxi inNewYorkCity is only 12 passengers per trip [6] Emptytaxis usually cruise along urban roads looking for customersThis is economically burdensome because they increase thetraffic volume which may lead to traffic congestion on the

HindawiJournal of Advanced TransportationVolume 2018 Article ID 8919721 13 pageshttpsdoiorg10115520188919721

2 Journal of Advanced Transportation

urban road network Overall the operational efficiency isrelatively low The main reason why customers decide tohire taxis is that they have limited time but heavy trafficprevents customers from reaching their destinations on time[7] Meanwhile the current system is unable to match theentire taxi demand during peak hoursMeasures to solve suchproblems are urgently needed

Accordingly this study was conducted with the objectiveof developing a shared autonomous taxi system (SATS) andinvestigating the efficiency of the system through simulationexperiments In the studied system customers are assumedto be willing to share a taxi with other customers and canrequest a taxi through a smartphone Shared autonomoustaxis (SATs) remain parked when unoccupied to reduce theroad burden A taxi is automatically assigned by the SATS topick a customer up after considering both unoccupied andoccupied taxis By taking advantage of unoccupied seats theSATS can reduce the total number of taxis required on theroad network

AVs are equipped with superior technological sensors[8] which are used to accurately perceive the surroundingenvironment and collect traffic information In this studyas probe vehicles SATs were deemed capable of collectingvaluable traffic information including data pertaining to thevehicle location and link travel time The collected trafficinformation can be used for path finding and categorizedas historical and real-time information The effectiveness ofusing the two types of information was investigated in thisstudy

The remainder of this paper is organized as follows InSection 2 the literature on taxi sharing and the performanceof shared AVs is reviewed The SATS considered in thisstudy is explained in Section 3 Section 4 presents the trafficinformation used for path finding in the SATS The simu-lation design and results are described in Section 5 Finallyconclusions and a discussion of future work are presented inSection 6

2 Literature Review

Autonomous taxi systems were proposed prior to the year2000 eg the autonomous dial-a-ride taxi system [9] eventhough the technology for navigation andpositioningwas notsufficiently advanced at that time With the development ofAVs a customer can request an SAT and ride it to hisherdestination The sharable taxis can be assigned to a singleperson or shared with other customers The deploymentof shared AVs (SAVs) can lead to a reduction in the totalnumber of private cars on urban road networks Fagnantand Kockelman [10] designed an agent-based model for SAVoperations in which four strategies are used to relocate AVswith the aim of minimizing waiting times for future travelersA 5-min interval is chosen as the iteration period At thebeginning of every 5-min interval the travel demand in everyzone is predetermined A parameter called ldquoblock balancerdquois proposed which represents the difference between theexpected demand and supply for SAVs in the upcoming5min By comparing the average number of trips servedby an SAV with that by a private car they concluded that

one SAV can replace approximately 11 conventional carsFagnant et al [11] provided more details with regard tolink-level travel times Their proposed model comprises foursubmodules SAV location and trip assignment SAV fleetgeneration SAV movement and SAV relocation An SAV isassigned first to the traveler who has been waiting for thelongest time Fagnant and Kockelman [12] developed SAVsimulations for clients with different origins and destinationsby considering dynamic ride sharing (DRS) Five conditionsare considered to judge whether a ride could be shared thetotal travel time and the increase in the remaining journeytime for riding passengers the increase in the total traveltime for a new passenger the possibility of the new passengerbeing picked up in the next 5min and the total travel timefor the two passengers Their experimental results suggestedthat DRS has the potential to reduce the total service time(which includes the waiting time) travel time and costfor users The authors also discussed the optimal SAV fleetsize from an economic viewpoint However they did notprovide delivery rules for the customers in a shared taxi Thisis important because these rules determine the remainingtime and travel time for each individual customer Burghoutet al [13] replaced private vehicles in Stockholm with theSAVs in a simulation study The sharing schemes includedpassengers with the same origin and destination the sameorigin but different destinations and different origins butthe same destination Their results indicated that only 5of the current number of private cars would be needed totransport commuters but the travel time increased by 13on average Lioris et al [14] suggested that if the detour timeincurred by serving a potential customer exceeds a maximaldetour time that customer should be rejected Levin et al[15] reported that when choosing between an occupied SAVand an unoccupied SAV the one that is able to arrive at thecustomer location first should be assigned to the customerThey further proposed that SAVs would increase congestionbecause of the additional trips made to reach each customerrsquosorigin It was found that the difference in the vehicle milestraveled between SAV scenarios and non-SAV scenarios wasprimarily due to the repositioning trips required to pick thenext passenger up

To optimize conventional taxi systems in terms of cus-tomer convenience and mitigation of traffic congestion amajor strategy has been to investigate dial-a-ride and sharedtaxi systems Studies on this strategy can be found in theliterature Teal [16] found that commuters could be the majorusers of carpool ride sharing which can reduce travel costsEven though ride sharing is generally applied to private carscommuters can call a taxi with the intention of sharing itDial [9] reported on an autonomous dial-a-ride taxi systemin which customers request a taxi via telephone and onlythe customer is involved in the process of requesting a rideassigning the trip scheduling the arrival and routing thevehicle The task of a driver is to simply follow instructionsprovided by the vehiclersquos computers The author also investi-gated an ideal autonomous taxi system that has the ability toassign a taxi to a customer in the shortest possible time Tao[17] used each customerrsquos choice for themaximum acceptablenumber of sharing customers and acceptable gender as inputs

Journal of Advanced Transportation 3

Path finding for private vehicle

Path variationTravel time variation

Searching the four types of SATPath finding for SAT

Demand generationDemand attribute

Demand generation

SAT assignment

Enrouteactions of

SAT

Road network

traffic flow

Figure 1 Shared autonomous taxi system (SATS) integrated into traffic flow

to an algorithm which then selects the taxi that is able toreach the customerrsquos location most rapidly to provide theservice Orey et al [18] proposed that a customer requestbe sent to all taxis or a subset of taxis and each taxi wouldrespond with its cost associated with the trip to the customerThe customer would then determine the acceptable lowestcost for taxi sharing and choose a taxi to use However acustomer finds it difficult to select one taxi from potentialtaxis Ota et al [19] studied a data-driven taxi ride-sharingsimulation in which taxis cruise the road network even whenempty A trip is assigned to the taxi that provides the lowestadditional cost if the passenger is assigned Ma et al [20]noted that a ride request made through a smartphone appshould be assigned to whichever taxi that minimizes theincrease in the travel distance resulting from the requestwhilemeeting the arrival time capacity and monetary constraintsof both the new passenger and the existing passenger(s)Nourinejad and Roorda [21] verified the performance ofride sharing by maximizing the savings on the total vehiclekilometers traveled andmaximizing thematching rate Najmiet al [22] investigated the effect of different dynamic sharingmethods on the performance of the ride-sharing system

Many previous studies have investigated SAV and taxisharing To correctly plan SAT paths traffic information isessential in the SATS However providing sufficient infor-mation to enable accurate path selection is difficult In thisstudy the link travel-time information collected by SATs wasanalyzed and the information applied to path finding for bothSATs and private vehicles

3 Shared Autonomous Taxi System

This section describes the components and implementationof an SATS This system is an improvement of the systemproposed in a previous study [23] The SATS is built on twoevents SAT assignment and enroute actions of the SAT andthe types of responses these actions warrant The proposedsystem assumes that all SATs can be shared by two requestsand these requests can only be made through a smartphoneIt should be noted that a request can contain more than onecustomer

This SATS operates on a network 119866 = (119873119860 119881119863)where 119873 is the set of nodes and 119860 is the set of links Thenetwork has a set of SATs 119881 that provides service to demand119863 The integration of this system with road network trafficflow is illustrated in Figure 1 The implementation steps aregrouped into four modules (1) demand generation (2) SATassignment (3) enroute actions of SAT and (4) road networktraffic flow The remainder of this section describes thesemodules in detail

31 Demand Generation The demand generation moduleintroduces customers to the proposed SATS In each timestep 120591 this module outputs a set of new customers whorequest an SAT The new customers and waiting customersare provided with the service during this time step and thewaiting customers are provided with this service before thenew customers

In this study we assumed that demand can be separatedinto single requests The origin and destination of eachrequest 119889 isin 119863 are denoted by 119874119889 and 119863119889 respectivelyFor ride-sharing problems departure time and arrival timeare key factors in improving matching rate [24] thereforeit was assumed that each request has an earliest departuretime denoted by119864119863119879119889 a latest arrival time denoted by 119871119860119879119889and travel-time flexibility denoted by 119879119879119865119889 [25] Request 119889should be served in the time window (119864119863119879119889 119871119860119879119889) [26]In this study 119864119863119879 was assumed to be the same as the SATrequest time Travel-time flexibility is the extra time that isacceptable to requests If the minimum travel time from 119874119889to 119863119889 is 119879(119874119889 119863119889) the travel-time flexibility of request 119889 is119879119879119865119889 = 119871119860119879119889 minus119864119863119879119889 minus119879(119874119889119863119889) ge 0 Once a request appearsat 120591 the SAT assignment event will be triggered

32 SAT Assignment The SAT assignment operates andresponds to the appearance of new demand This moduledescribes the specific logic used to assign SATs to demand inthe SATS

The SATS assumes the existence of a virtual centralcontrol system that knows the status of all SATs and requeststhis system can also assign an SAT to each request and selectthe path for the SAT The output of the SAT assignment is


21 3 4 5

7 6 8 9 10

1211 13 14 15

1716 18 19 20

2221 23 24 25

Od

Dd

Currently empty SAT

Expected-empty SAT

Expected-sharable SAT

Passed link

Remaining link

Currently sharable SAT

Origin of a trip

Destination of a trip

Figure 2 Four types of available SATs

an update of the SAT status including unoccupied single-rider occupied shared and change-of-request status (iein-service or waiting to be served) In this study a ridercorresponds to a request Each SAT is either parked at a nodeor serving a customer at any given time therefore in thisstudy it was assumed that nodes have (infinite) parking space

321 Four Types of Available SATs In the SATS once anew request is generated the virtual central control systemsearches the available SATs for the request All the searchedSATs should be located within the distance 119863119898119886119909 from therequest where 119863119898119886119909 is the radius of the search area Thereare four types of available SATs according to their statusand possibility of sharing currently empty expected-emptycurrently sharable and expected-sharable SATs

A currently empty SAT is an unoccupied SAT As shownin Figure 2 when request 119889 occurs at node 119874119889 the virtualcentral control system first tries to check whether a currentlyempty SAT is available In this figure an unoccupied SAT isparked at node 15

An expected-empty SAT is a taxi that is occupied by othercustomer(s) presently and the rider will get off within themaximum waiting time 119879119898119886119909 Next the SAT can travel to 119889As shown in Figure 2 the origin and destination of rider inan expected-empty SAT are node 6 and node 11 respectivelyand the taxi is traveling on the planned path In this case this

SAT can first proceed to node 11 to drop the rider off and thentravel to 119874119889 to pick 119889 up

A currently sharable SAT is defined as a single-rideroccupied taxi that can be shared with another request if therequirements for sharing can be met which are threefoldThe first requirement is that the origin of the new request 119874119889should be on a path in the path set of the rider 119888 The pathset of each OD pair is predetermined using a 119896-shortest pathalgorithm [27 28]The 119896-shortest path algorithm canwork inthis study when the size of a network is not large In additionits application can reduce the time for processing everyrequest since candidate SATs for sharing can be screenedeffectively The second requirement is that the destination119863119888of rider 119888 should be on a path in the path set of the newrequest or the destination of the new request 119863119889 should beon a path in the path set of rider 119888 The third requirement isthat both rider 119888 and request 119889 can arrive at their destinationswithin 119871119860119879119888 and 119871119860119879119889 respectively This is called sharingcheck in this SATS As shown in Figure 2 the origin anddestination of the rider are node 4 and node 24 respectivelyThe taxi may change its route from the original path (4 997888rarr9 997888rarr 14 997888rarr 19 997888rarr 24) at node 14 to pick request 119889 up andthe two trips will share this SAT on the link from 13 to 24

An expected-sharable SAT is an SAT that is presentlyoccupied by two riders and can be shared by request 119889 andthe remaining rider after one rider gets off the vehicle For thistype of SAT the requirements for sharing are also threefoldThe first two requirements are the same as the first tworequirements for currently sharable SAT and rider 119888 in thesharing requirements is the rider who can arrive at theirdestination later than the other rider The last requirementis that both the rider and request 119889 can arrive at theirdestinations within their respective latest arrival time Asshown in Figure 2 an expected-sharable SAT is currentlyoccupied by two riders After one rider arrives at theirdestination (node 8) the SAT can be shared by the remainingrider and request 119889

In this SATS a currently sharable SAT is a single-rideroccupied SAT and a single-rider occupied SAT can alsopossibly be an expected-empty SAT A two-rider occupiedSAT can be an expected-empty SAT or an expected-sharableSAT For searched occupied SAT the virtual central controlsystem checks the possibility of this SAT being a currentlysharable or an expected-sharable SAT because the path setfor eachODpair is found using the 119896-shortest path algorithmand customers are prohibited from traveling on paths beyondthe path set

322 Path Finding for SAT Before SAT assignment thevirtual central control system searches for a potential pathfor each of the four types of SATs using the shortest travel-time path algorithm When an SAT is finally assigned theSAT transports customers on the potential path The path ofa traveling SAT cannot be changed until customers arrive attheir destination or the SAT is shared with a new requestFor SATs that are not parked at 119874119889 at the moment that pathfinding begins to correctly find the shortest travel-time paththe mean travel time for each possible path should includethe remaining part of the ongoing path Figure 3 takes the


Time segments

Links

+

+

+

+

1ℎ

+ ge +

1ℎ+2Δℎ

1ℎ+3Δℎ

1ℎ+4Δℎ

2ℎ 3ℎ 4ℎ 5ℎ mℎ

3ℎ+Δℎ 4ℎ+Δℎ 5ℎ+Δℎ mℎ+Δℎ

+ ge + 4ℎ+2Δℎ 5ℎ+2Δℎ mℎ+2Δℎ

+ ge + 5ℎ+3Δℎ mℎ+3Δℎ

+ ge + mℎ+4Δℎ

+ minus ge +

Path o Path p Path q

54

Figure 3 Path travel time of expected-empty taxi

expected-empty SAT as an example Information in this tableis the mean travel time of each link in each time segmentIn the figure suppose that path 119900 is the ongoing path of anexpected-empty SAT and links 2 and 3 are the remaininglinksThe SAT travels to119874119889 through path119901 Path 119902 is one pathfrom the origin of 119889 to the destinationThe travel time for thisnew path is the sum of the travel time for the remaining linksof paths 119900 119901 and 119902 The consideration of the transition in thetraffic condition is explained in Section 4

323 SAT Assignment and Status Update To reduce the totaltravel time a request is served by the SAT with the minimumtravel time to its destination The virtual central controlsystem executes the following steps sequentially

Step 1 The central control system searches for currentlyempty SATs within 119863119898119886119909 from the origin of request 119889 andthen searches for the minimum travel-time paths for eachsearched SATThe available path set of every currently emptySAT is the combination of two path sets the path set fromthe SATrsquos location to 119874119889 and the path set from 119874119889 to 119863119889The consideration of the combination of two path sets isreasonable because our aim is to ensure that the requestarrives at the destination in the minimum travel time whentime transition is considered which is described in Section 4The minimum travel times with each searched currentlyempty SAT can be recorded as 1198791198881198901 1198791198881198902 1198791198881198901198991 where1198991 is the number of searched currently empty SATs Theminimum travel time is recorded as 119879119888119890119898119894119899Step 2 The system searches for expected-empty SATs Theshortest travel-time path for each expected-empty SAT issearched as described in Section 322 The minimum traveltimes of each expected-empty SAT are 1198791198901198901 1198791198901198902 1198791198901198901198992where 1198992 is the number of expected-empty taxis The mini-mum travel time is recorded as 119879119890119890119898119894119899Step 3 The system searches for currently sharable andexpected-sharable SATs The central control system firstchecks whether the searched SATs meet the requirements

for sharing The searched SATs are feasible SATs if therequirements aremet and a potential path is assigned to eachfeasible SAT The minimum travel times for the new requestcan be recorded as 1198791198881199041 1198791198881199042 1198791198881199041198993 with all feasiblecurrently sharable SATs and as 1198791198901199041 1198791198901199042 1198791198901199041198994 withall feasible expected-sharable SATs where 1198993 and 1198994 are thenumber of feasible currently sharable and expected-sharableSATs The minimum travel time with currently sharable SATand that with expected-sharable SAT are 119879119888119904119898119894119899 and 119879119890119904119898119894119899respectively

Step 4 The SAT with the minimum travel time 119879119898119894119899 isassigned to the request eventually

Step 5 If the virtual central control system cannot find 119879119898119894119899which means that no SAT is available the request 119889 is addedto the waiting list

The SAT assignment procedure is shown in Figure 4After the assignment of the SAT the status of both the requestand SAT will be updated The completion of this moduletriggers the third module enroute actions of the SAT

33 Enroute Actions of SAT When an SAT departs from theorigin of a request themodule of enroute actions is triggeredIn this process the SAT takes the customer to the destinationThe predicted travel time of the SAT varies with changingtraffic conditionsThe status and planned path of the SAT alsochangewhen the SATproceeds to the destination orwhen theSAT is shared with a new request

331 Status Variation of SAT and Its Causes It is possible thatthe status of an SAT changes to an expected-empty SAT froma single-rider occupied SATor shared SATThis is because theSAT is selected to serve the next request It is possible that thepick-up time for the next request is delayed or early accordingto variations in traffic conditions In this study we assumethat once an SAT is assigned to a request the request cannotrefuse the assigned SATThe SAT travels to the destination ofthe last rider and parks at the destination if it is not selected


Travel time for ongoing path

Demand

Waiting listNo

C-E SAT Currently empty SAT C-S SAT Currently sharable SATE-E SAT Expected-empty SAT E-S SAT Expected-sharable SAT

Path combination for C-E SAT Sharing check

Path combination for E-E SAT

Shortest path for C-S SAT and E-S SAT

Shortest path for C-E SAT

Shortest path for E-E SAT

Teemin among E-E SATsTcemin among C-E SATs

Tcsmin and T esmin amongC-S SATs and E-S SATs

Tmin is found

E-E SATNo

YesC-S SAT

NoYes

Demand can be servedNo

YesE-S SAT

YesC-E SAT

４ＧＣＨ = ４＝ＭＧＣＨ

４ＧＣＨ = ４ＭＧＣＨ

４ＧＣＨ = ４＝ＧＣＨ

Figure 4 SAT assignment procedure

to serve other requests soonThe status of the SAT is updatedto unoccupied from single-rider occupied or shared

The departed taxi is a single-rider occupied SATor sharedSAT On the way to the destination the status of the taxi canchange from one of the two above statuses to another statusFor a shared SAT the status changes to single-rider occupiedwhen one rider reaches the destination The SAT can beshared with following requests which results in SAT tripchaining [29] The status of the SAT is still shared if it picks anew request up right at the node where one rider exits If thedeparted SAT is only occupied by one rider it can be sharedwith the following request on the way to the destination Inthis study two types of sharing are considered these typesdetermine the delivery sequence of the two riders and aredescribed in Section 332 A new path is also necessary tobe searched for picking up and dropping off customers thisaspect is described in Section 333

332 Sharing Types and Customer-Delivery Rule Based onthe location of the origin and destination of the rider andthose of the new request sharing is divided into two typesextended-path sharing (EP sharing) and in-path sharing (IPsharing) EP sharing means that sharing extends the plannedpath of the taxi because the destination of the new request isnot on any path in the path set of the rider as shown in Figures5(a) and 5(b) In IP sharing both the origin and destinationof the new request are on the 119896-shortest paths of the rider asshown in Figures 5(c) 5(d) 5(e) and 5(f) Even if the originand destination of the new request are located on differentpaths of the path set of the rider the case can be consideredprovided there is a common node where SATs can detour toa new path to pick up or drop off a rider from the plannedpath

In EP sharing the first-in-first-out (FIFO) order is used asthe customer-delivery rule that is the rider is delivered first


21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(a)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(b)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(c)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(d)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(e)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(f)

Figure 5 Sharing types in SATS

In IP sharing the new request can arrive at the destinationearlier than or at the same time as the rider It is necessaryto determine the sharing type and delivery rule becauseit determines the minimum total travel time for the twotrips

333 Shared Path for Currently Sharable SAT and Expected-Sharable SAT If an SAT meets the first two sharing require-ments described in Section 323 the central control systemfinds a new path for each feasible sharable SAT as the sharedpath The path set of this SAT is the combination of thepath sets of the rider 119888 and the new request 119889 The methodmentioned in Section 322 is used to select paths that canmake the rider 119888 arrive at the destination within 119871119860119879119888The travel times for these paths for the new request are119879ss1 119879ss2 119879s1199041198995 and theminimum time is119879ssmin In thiscase the path with119879ssmin is assigned as the shared path to thetaxiThe SAT travels on the shared path if it is finally selectedto provide service to the new request

The above three modules constitute the implementationof the SATS which is integrated to the traffic system

34 Road Network Traffic Flow In the simulation it wasassumed that road network traffic flow consists of taxis andprivate vehicles The path of a private vehicle was selectedstatistically Considering the diverse preferences of privatevehicle drivers the path choice probability was calculated

by using the path size logit model [30ndash32] The model isformulated as follows [32]

119875( 119894119862119899) =119875119878119894119899119890119881119894119899

sum119895isin119862119899 119875119878119895119899119890119881119895119899(1)

119875119878119894119899 = sum119886isinΓ119894

( 119897119886119871 119894)1

sum119895isin119862119899 (119871 119894119871119895) 120575119886119895 (2)

where 119881119894119899 is the systematic term of utility of path 119894 for user 119899119862119899 is the path set 119875119878119894119899 is the size of path 119894 119897119886 is the length oflink 119886 119871 119894 is the length of path 119894 120575119886119895 is 1 if link 119886 is in path 119895and 0 otherwise and Γ119894 is the set of links of path 1198944 Travel-Time Information

Like probe vehicles traveling SATs can also collect trafficinformation including data pertaining to the vehicle locationand link travel time The collected travel-time informationcan be used for path finding by both SATs and privatevehicles Link travel velocities are assumed to be calculatedas V119886 = 161 lowast ln(215119896119886) [33] where 215 vehiclesmile arethe density of the traffic jam and 119896119886 is the density of link119886 The link density is updated every minute The travel timefor an SAT or a private vehicle on link 119886 is assumed to berandom and is normally distributed with 119905119886 sim 119873(120583119886 1205902119886)where 120583119886 = (119897119886V119886) and 1205902119886 = 120572(119897119886V119886)2 in which 120583119886 and1205902119886 are the population mean value and variance of the link


Links

Time segments

1

+

1ℎ 2ℎ 3ℎ aℎ mℎ

(a) Case of real-time traffic information

Time segments

Links


+ ge + 3ℎ+Δℎ aℎ+Δℎ mℎ+Δℎ

+ ge + aℎ+2Δℎ mℎ+2Δℎ

1ℎ+3Δℎ + minus ge + aℎ+3Δℎ mℎ+3Δℎ

1ℎ+4Δℎ + ge + mℎ+4Δℎ

+ minus ge +

+

+

+

+ 4

(b) Case of historical traffic information

Figure 6 Use of two types of traffic information in SATS

travel time on link 119886 respectively [34] In the proposed SATSit is assumed that the link travel times are independent Themean value of the path travel time can be expressed by 120583119901 =sum119886isin119901 120583119886 where 120583119886 is the expected value of the travel time(119905119886) experienced by SATs on link 119886 Considering the onlineinformation paradox [35] which means that the provisionof incomplete or inaccurate information could deterioratethe transportation system it is urgent and necessary toprovide accurate information and use the information inan appropriate manner In this study two methodologiesfor using the collected information real-time informationand historical information were investigated However thisstudy has the limitation that no exogenous disturbance wasconsidered The reason for this lack of consideration is thatthe traffic environment is expected to becomemore stable andtraffic safety can be improved in the autonomous vehiculartransportation system

In the case of real-time traffic information the informa-tion collected by SATs at time segment ℎ is used for pathfinding at ℎ + Δℎ That is the link travel-time informationfor link 119886 at ℎ + Δℎ is 120583119886ℎ The information update interval isΔℎ As shown in Figure 6(a) if the path of an SAT at ℎ + Δℎcontains 119898 links the travel-time information for the pathwillbe the sum of the 120583119886ℎ for the119898 links This is the same for thecurrent travel-time system

In the case of historical traffic information a traffic infor-mation database is first established as shown in Figure 6(b)The information in the database is the mean value of the linktravel time experienced by SATs in every time segment untilthe previous day Through traffic information collection andaccumulation information in the database is updated eachday This information is utilized for path finding in the same

time segment in the following days In this methodologythe transition of traffic condition is also considered [36] Asshown in Figure 6(b) the departure time of an SAT is ℎ0which is greater than ℎ and less than ℎ + Δℎ The total traveltime for the first 119886 links is ℎ119886 The traffic information for link3 at ℎ + Δℎ is used for path finding because the total traveltime for the first two links exceeds ℎ+Δℎminusℎ0 and is less thanℎ + 2Δℎ minus ℎ05 Simulation Design and Results

To determine the effect of traffic information and howdifferent SAT fleet sizes perform in the SATS in this studyseveral sets of simulation experiments were conducted inMATLAB on an Intel Core CPU running at 300GHz

The experiments were performed on the Sioux Fallsnetwork as shown in Figure 7 The network had 24 nodes76 links and 552 OD pairs The metric used in the 119896-shortestpath algorithm was distance and 119896 was set 10 The free-flowspeed of all the links was set to 40 kmh The travel time fora link was calculated using its length and the speed obtainedusing the equation in Section 4 In this study 120572was set to 021[37 38]

The trip demand by taxi requests and private vehicle usersis shown in Figure 8 The demand was generated from aPoisson distribution every minute and spread over a 24-hperiod based on the temporal distribution of US NHTS trip-start rates in 2009 [39]The origin and destination of each tripwere generated randomly

Each experimentwas simulated for 80 days Initially SATswere distributed evenly on every node in the network Weassumed that all SATs could be relocated at 000 am to


6

89 7

1610 1812 11

17

191514

2223

202113 24

1

3 4 5

23 km

2 km

4 km

14 km

41 km

2 km

14 km2 km

Length of links notshown in the figure is 1 km

Figure 7 Sioux Falls network

handle the demand of the next day119863119898119886119909 was set as 10 km and119879119898119886119909 was set as 10minThe latest arrival time of a request wasthe sum of the travel time from its origin to the destinationat the average travel speed (20 kmh) and an acceptable timethat follows 119880(0 10) Δℎ was set as 5min in all the scenariosBoth on the first simulation day of the historical trafficinformation-based scenario and in the first time segment ofthe real-time traffic information-based scenario the travel-time information used for path finding was considered tohave been obtained under free flow The simulation resultsare shown in Figures 9ndash11

51 Effect of Traffic Information Figure 9(a) presents the day-to-day change in the mean travel time for PV users andSAT requests with real-time traffic information Figure 9(b)illustrates the travel times with historical traffic informationand Figure 9(c) shows the mean waiting times for SATrequests in the case where the SAT fleet size was 30 of thefixed peak hour demand of SAT requests

All the travel times and the waiting time in the historicaltraffic information-based scenario decreased gradually asthe number of simulation days increased and the timesconverged on approximately the 30th simulation day Thetravel times and waiting time in the real-time information-based scenario were stable at a larger time level during theentire simulation All the travel times at the stable levelin the historical traffic information-based scenario had lessfluctuation than those in the real-time traffic information-based scenario In the historical information-based scenariothe times decreased because with the collection and accu-mulation of traffic information and the consideration of timetransition the central control center can find amore accurate

shortest travel-time path for each SAT compared to thepast simulation days and the real-time traffic information-based scenario These results confirm that historical trafficinformation is better than real-time traffic information forpath finding in the proposed SATS This is consistent withthe conclusions in the literature [38]

In both scenarios the total travel times for taxi requestsare greater than those for private vehicle users This isreasonable because the total travel time includes the waitingtime until the SATrsquos arrival The in-vehicle travel times fortaxi requests are smaller than those for private vehicle usersbecause the probabilistic path choice behavior of privatevehicles is considered and some private vehicles do not selectthe shortest travel time path

52 Effect of SAT Fleet Size Figures 10 and 11 present the day-to-day changes in the total travel times and waiting times forthe two traffic information source cases The SAT fleet sizewas set to 30 40 and 50 of the fixed peak hour demandof SAT requests

The figures clearly show that the total travel time for thetaxi requests in both scenarios decreases as the taxi fleet sizeincreasesThe difference among the different fleet sizes is dueto the difference in the waiting time The in-vehicle traveltime does not vary significantly among different fleet sizesThat is as the supply of SATs increases the number of taxirequestswhohave towait for taxis decreases In particular thedifference between the cases with 30 and 40 of fleet sizesis considerably greater than the difference between the caseswith 40 and 50 of fleet sizes These results demonstratethat an insufficient fleet size in the SATS would nonlinearlyworsen the service level of the SATS

In addition Figure 11(b) indicates that as the number oftaxis increases convergence occurs earlier in the historicalinformation-based scenario This is because in the case witha larger fleet size the statistical reliability of the travel-time information accumulated in the historical informationdatabase can become higher because of the larger amountof collected data Accordingly the accuracy for searching foravailable SATs and the accuracy of path findings can increase

6 Conclusions and Future Work

This paper proposed a framework for an SATS that utilizescollected travel-time information to reduce taxi demand andincrease the possibility of reducing the travel time for taxicustomers In the proposed SATS framework in which thereare no drivers SATs utilize collected travel-time informationfor path finding The use of two types of information wasinvestigated historical information and real-time informa-tion The results of the simulation experiments conducted inthis study verify that using the two types of information iseffective

More specifically the simulations demonstrated the effectof using historical travel-time information for path findingand reducing the travel time in the SATS In particular as thenumber of simulation days increased the customers in theSATSwith the historical traffic information obtained increas-ingly smaller travel times until the travel time converged


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24Start Hour

Private vehicle userTaxi request

0

1000

2000

3000

4000

5000

Dem

and

(Tri

ps)

Figure 8 Trip demand by taxi requests and private vehicle users

Total travel time for SAT request with RITravel time for PV user with RIIn-vehicle travel time for SAT request with RI

105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day

(a) Travel time with real-time traffic information

Total travel time for SAT request with HITravel time for PV user with HIIn-vehicle travel time for SAT request with HI

105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day

(b) Travel time with historical traffic information

0 10 20 30 40 50 60 70 80Simulation day

Waiting time for SAT request with RIWaiting time for SAT request with HI

085

09

095

1

105

11

115

12

125

13

135

Tim

e (m

in)

(c) Waiting time

Figure 9 Day-to-day changes in (a) travel time with RI (b) travel time with HI and (c) waiting time for SAT request and PV user (RIreal-time information HI historical information)


SAT fleet size is 30 of the peak hour demandSAT fleet size is 40 of the peak hour demandSAT fleet size is 50 of the peak hour demand

10 20 30 40 50 60 70 800Simulation day

105

11

115

12

125

13

135

Tim

e (m

in)

(a) Total travel time in the real-time traffic information-based scenario

0 10 20 30 40 50 60 70 80Simulation day


105

11

115

12

125

13

135

Tim

e (m

in)

(b) Total travel time in the historical traffic information-based scenario

Figure 10 Day-to-day changes in total travel time for taxi requests in SATS with several SAT fleet sizes


03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day

(a) Waiting time in real-time traffic information-based scenario


03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day

(b) Waiting time in historical traffic information-based scenario

Figure 11 Day-to-day changes in waiting time for taxi requests in SATS with several SAT fleet sizes

to a steady level The stable travel times were also smallerthan those in the real-time traffic information-based systemwhere the travel times fluctuated slightly around the constantvalues The study results also indicate that insufficient fleetsize in the SATS would nonlinearly worsen the service levelof the SATS and that as the number of SATs increasesconvergence occurs earlier in the historical information-based scenario

Future study on this topic can focus on the relocation pro-cess of SATs Routing strategies for SATs and information useunder exogenous disturbances also need to be studied Thevariation in SAT demand to the different service level should

be considered in our future research This consideration willprovide more realistic process until the convergence of trafficstates Another area of future research in order to offer a betterlevel of service is examination of the effect of reassigninga new taxi to a request when the pick-up time of theassigned taxi increases owing to variations in traffic condi-tions

Conflicts of Interest

The authors declare that there are no conflicts of interestregarding the publication of this paper


Acknowledgments

This work was supported by JSPS KAKENHI Grant nos16H02367 and 16K12825 The first author would like to thankthe China Scholarship Council (CSC) for financial support

References

[1] S Horl F Ciari and KW Axhausen Recent Perspectives onTheImpact of Autonomous Vehicles Institute for Transport Planningand Systems 2016

[2] J Patel ldquonuTonomy beats Uber to launch first self-driving taxirdquo2016 httpwwwautoblogcom

[3] H Chai ldquo4 Self-Driving Buses Tested in Shenzhenrdquo httpwwwchinadailycomcnchina201

[4] R Krueger T H Rashidi and J M Rose ldquoPreferences forshared autonomous vehiclesrdquo Transportation Research Part CEmerging Technologies vol 69 pp 343ndash355 2016

[5] P Bansal K M Kockelman and A Singh ldquoAssessing publicopinions of and interest in new vehicle technologies AnAustin perspectiverdquo Transportation Research Part C EmergingTechnologies vol 67 pp 1ndash14 2016

[6] M Bloomberg and D Yassky ldquoTaxicab Fact Bookrdquo httpwwwnycgovhtml

[7] WWuW S Ng S Krishnaswamy andA Sinha ldquoTo taxi or notto taxi - Enabling personalised and real-time transportationdecisions for mobile usersrdquo in Proceedings of the 2012 IEEE 13thInternational Conference on Mobile Data Management MDM2012 pp 320ndash323 India July 2012

[8] Z Jia A Balasuriya and S Challa ldquoSensor fusion-basedvisual target tracking for autonomous vehicles with the out-of-sequencemeasurements solutionrdquoRobotics andAutonomousSystems vol 56 no 2 pp 157ndash176 2008

[9] R B Dial ldquoAutonomous dial-a-ride transit introductoryoverviewrdquo Transportation Research Part C Emerging Technolo-gies vol 3 no 5 pp 261ndash275 1995

[10] D J Fagnant and K M Kockelman ldquoThe travel and envi-ronmental implications of shared autonomous vehicles usingagent-based model scenariosrdquo Transportation Research Part CEmerging Technologies vol 40 pp 1ndash13 2014

[11] D J Fagnant K M Kockelman and P Bansal ldquoOperationsof shared autonomous vehicle fleet for Austin Texas marketrdquoTransportation Research Record vol 2536 pp 98ndash106 2015

[12] D J Fagnant and K M Kockelman ldquoDynamic ride-sharingand fleet sizing for a system of shared autonomous vehicles inAustin Texasrdquo Transportation pp 1ndash16 2016

[13] W Burghout P J Rigole and I Andreasson ldquoImpacts of sharedautonomous taxis in a metropolitan areardquo in Proceedings ofthe 94th Annual Meeting of the Transportation Research BoardWashington wash USA 2015

[14] E Lioris G Cohen andA de La Fortelle ldquoEvaluation of Collec-tive Taxi Systems by Discrete-Event Simulationrdquo in Proceedingsof the 2010 Second International Conference on Advances inSystem Simulation (SIMUL) pp 34ndash39 Nice France August2010

[15] M W Levin T Li and S D Boyles ldquoA general frameworkfor modeling shared autonomous vehiclesrdquo in Proceedings ofthe 94th Annual Meeting of the Transportation Research BoardWashington D 2016

[16] R F Teal ldquoCarpooling who how and whyrdquo TransportationResearch Part A General vol 21 no 3 pp 203ndash214 1987

[17] C-C Tao ldquoDynamic taxi-sharing service using intelligenttransportation system technologiesrdquo in Proceedings of the Inter-national Conference on Wireless Communications NetworkingandMobile Computing (WiCOM rsquo07) pp 3204ndash3207 September2007

[18] P M DrsquoOrey R Fernandes and M Ferreira ldquoEmpirical eval-uation of a dynamic and distributed taxi-sharing systemrdquo inProceedings of the 2012 15th International IEEE Conference onIntelligent Transportation Systems ITSC 2012 pp 140ndash146 USASeptember 2012

[19] M Ota H Vo C Silva and J Freire ldquoA scalable approach fordata-driven taxi ride-sharing simulationrdquo in Proceedings of the3rd IEEE International Conference on Big Data IEEE Big Data2015 pp 888ndash897 USA November 2015

[20] S Ma Y Zheng and O Wolfson ldquoReal-Time City-ScaleTaxi Ridesharingrdquo IEEE Transactions on Knowledge and DataEngineering vol 27 no 7 pp 1782ndash1795 2015

[21] M Nourinejad and M J Roorda ldquoAgent based model fordynamic ridesharingrdquo Transportation Research Part C Emerg-ing Technologies vol 64 pp 117ndash132 2016

[22] A Najmi D Rey and T H Rashidi ldquoNovel dynamic for-mulations for real-time ride-sharing systemsrdquo TransportationResearch Part E Logistics and Transportation Review vol 108pp 122ndash140 2017

[23] Z Liu T Miwa W Zeng and T Morikawa ldquoAn agent-basedsimulation model for shared autonomous taxi systemrdquo AsianTransport Studies vol 5 no 1 pp 1ndash13 2018

[24] M Stiglic N Agatz M Savelsbergh and M Gradisar ldquoMakingdynamic ride-sharing work The impact of driver and riderflexibilityrdquo Transportation Research Part E Logistics and Trans-portation Review vol 91 pp 190ndash207 2016

[25] H Hosni J Naoum-Sawaya and H Artail ldquoThe shared-taxiproblem Formulation and solution methodsrdquo TransportationResearch Part B Methodological vol 70 pp 303ndash318 2014

[26] J Y Yen ldquoFinding the K shortest loopless paths in a networkrdquoManagement Science vol 17 pp 712ndash716 19707119

[27] Y Molenbruch K Braekers and A Caris ldquoTypology andliterature review for dial-a-ride problemsrdquoAnnals of OperationsResearch vol 259 no 1-2 pp 295ndash325 2017

[28] W Zeng T Miwa and T Morikawa ldquoApplication of thesupport vectormachine and heuristic k-shortest path algorithmto determine the most eco-friendly path with a travel timeconstraintrdquo Transportation Research Part D Transport andEnvironment vol 57 pp 458ndash473 2017

[29] M W Levin K M Kockelman S D Boyles and T Li ldquoAgeneral framework for modeling shared autonomous vehi-cles with dynamic network-loading and dynamic ride-sharingapplicationrdquo Computers Environment and Urban Systems vol64 pp 373ndash383 2017

[30] E Frejinger M Bierlaire and M Ben-Akiva ldquoSampling ofalternatives for route choicemodelingrdquoTransportation ResearchPart B Methodological vol 43 no 10 pp 984ndash994 2009

[31] D Li T Miwa and T Morikawa ldquoDynamic route choicebehavior analysis considering en route learning and choicesrdquoTransportation Research Record no 2383 pp 1ndash9 2013

[32] S Bekhor M Ben-Akiva and M S Ramming ldquoAdaptation ofLogit Kernel to route choice situationrdquo Transportation ResearchRecord vol 1805 no 1 pp 78ndash85 2002

[33] H Greenberg ldquoAn analysis of traffic flowrdquo Operations Researchvol 7 pp 79ndash85 1959


[34] R Seshadri and K K Srinivasan ldquoAlgorithm for determiningmost reliable travel time path on network with normallydistributed and correlated link travel timesrdquo TransportationResearch Record no 2196 pp 83ndash92 2010

[35] K PWijayaratna VVDixit D-B Laurent and S TWaller ldquoAnexperimental study of the Online Information Paradox Doesen-route information improve road network performancerdquoPLoS ONE vol 12 no 9 Article ID e0184191 2017

[36] T Morikawa and T Miwa ldquoPreliminary analysis on dynamicroute choice behavior Using probe-vehicle datardquo Journal ofAdvanced Transportation vol 40 no 2 pp 141ndash163 2006

[37] T Yamamoto T Miwa T Takeshita and T Morikawa ldquoUpdat-ing dynamic origin-destination matrices using observed linktravel speed by probe vehiclesrdquo Transportation and TrafficTheory pp 723ndash738 2009

[38] T Miwa and M G Bell ldquoEfficiency of routing and schedulingsystem for small and medium size enterprises utilizing vehiclelocation datardquo Journal of Intelligent Transportation SystemsTechnology Planning andOperations vol 21 no 3 pp 239ndash2502016

[39] Federal Highway Administration National Household TravelSurvey US Department of Transportation Washington washUSA 2009 httpnhtsornlgov2009pubsttpdf

International Journal of

AerospaceEngineeringHindawiwwwhindawicom Volume 2018

RoboticsJournal of

Hindawiwwwhindawicom Volume 2018


Active and Passive Electronic Components

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics

Hindawiwwwhindawicom

Volume 2018

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Control Scienceand Engineering

Journal of



Journal ofEngineeringVolume 2018

SensorsJournal of



RotatingMachinery


Modelling ampSimulationin EngineeringHindawiwwwhindawicom Volume 2018


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation


Hindawi

wwwhindawicom Volume 2018

Advances in

Multimedia

Submit your manuscripts atwwwhindawicom


urban road network Overall the operational efficiency isrelatively low The main reason why customers decide tohire taxis is that they have limited time but heavy trafficprevents customers from reaching their destinations on time[7] Meanwhile the current system is unable to match theentire taxi demand during peak hoursMeasures to solve suchproblems are urgently needed

Accordingly this study was conducted with the objectiveof developing a shared autonomous taxi system (SATS) andinvestigating the efficiency of the system through simulationexperiments In the studied system customers are assumedto be willing to share a taxi with other customers and canrequest a taxi through a smartphone Shared autonomoustaxis (SATs) remain parked when unoccupied to reduce theroad burden A taxi is automatically assigned by the SATS topick a customer up after considering both unoccupied andoccupied taxis By taking advantage of unoccupied seats theSATS can reduce the total number of taxis required on theroad network

AVs are equipped with superior technological sensors[8] which are used to accurately perceive the surroundingenvironment and collect traffic information In this studyas probe vehicles SATs were deemed capable of collectingvaluable traffic information including data pertaining to thevehicle location and link travel time The collected trafficinformation can be used for path finding and categorizedas historical and real-time information The effectiveness ofusing the two types of information was investigated in thisstudy

The remainder of this paper is organized as follows InSection 2 the literature on taxi sharing and the performanceof shared AVs is reviewed The SATS considered in thisstudy is explained in Section 3 Section 4 presents the trafficinformation used for path finding in the SATS The simu-lation design and results are described in Section 5 Finallyconclusions and a discussion of future work are presented inSection 6

2 Literature Review

Autonomous taxi systems were proposed prior to the year2000 eg the autonomous dial-a-ride taxi system [9] eventhough the technology for navigation andpositioningwas notsufficiently advanced at that time With the development ofAVs a customer can request an SAT and ride it to hisherdestination The sharable taxis can be assigned to a singleperson or shared with other customers The deploymentof shared AVs (SAVs) can lead to a reduction in the totalnumber of private cars on urban road networks Fagnantand Kockelman [10] designed an agent-based model for SAVoperations in which four strategies are used to relocate AVswith the aim of minimizing waiting times for future travelersA 5-min interval is chosen as the iteration period At thebeginning of every 5-min interval the travel demand in everyzone is predetermined A parameter called ldquoblock balancerdquois proposed which represents the difference between theexpected demand and supply for SAVs in the upcoming5min By comparing the average number of trips servedby an SAV with that by a private car they concluded that

one SAV can replace approximately 11 conventional carsFagnant et al [11] provided more details with regard tolink-level travel times Their proposed model comprises foursubmodules SAV location and trip assignment SAV fleetgeneration SAV movement and SAV relocation An SAV isassigned first to the traveler who has been waiting for thelongest time Fagnant and Kockelman [12] developed SAVsimulations for clients with different origins and destinationsby considering dynamic ride sharing (DRS) Five conditionsare considered to judge whether a ride could be shared thetotal travel time and the increase in the remaining journeytime for riding passengers the increase in the total traveltime for a new passenger the possibility of the new passengerbeing picked up in the next 5min and the total travel timefor the two passengers Their experimental results suggestedthat DRS has the potential to reduce the total service time(which includes the waiting time) travel time and costfor users The authors also discussed the optimal SAV fleetsize from an economic viewpoint However they did notprovide delivery rules for the customers in a shared taxi Thisis important because these rules determine the remainingtime and travel time for each individual customer Burghoutet al [13] replaced private vehicles in Stockholm with theSAVs in a simulation study The sharing schemes includedpassengers with the same origin and destination the sameorigin but different destinations and different origins butthe same destination Their results indicated that only 5of the current number of private cars would be needed totransport commuters but the travel time increased by 13on average Lioris et al [14] suggested that if the detour timeincurred by serving a potential customer exceeds a maximaldetour time that customer should be rejected Levin et al[15] reported that when choosing between an occupied SAVand an unoccupied SAV the one that is able to arrive at thecustomer location first should be assigned to the customerThey further proposed that SAVs would increase congestionbecause of the additional trips made to reach each customerrsquosorigin It was found that the difference in the vehicle milestraveled between SAV scenarios and non-SAV scenarios wasprimarily due to the repositioning trips required to pick thenext passenger up

To optimize conventional taxi systems in terms of cus-tomer convenience and mitigation of traffic congestion amajor strategy has been to investigate dial-a-ride and sharedtaxi systems Studies on this strategy can be found in theliterature Teal [16] found that commuters could be the majorusers of carpool ride sharing which can reduce travel costsEven though ride sharing is generally applied to private carscommuters can call a taxi with the intention of sharing itDial [9] reported on an autonomous dial-a-ride taxi systemin which customers request a taxi via telephone and onlythe customer is involved in the process of requesting a rideassigning the trip scheduling the arrival and routing thevehicle The task of a driver is to simply follow instructionsprovided by the vehiclersquos computers The author also investi-gated an ideal autonomous taxi system that has the ability toassign a taxi to a customer in the shortest possible time Tao[17] used each customerrsquos choice for themaximum acceptablenumber of sharing customers and acceptable gender as inputs






Demand generation

SAT assignment

Enrouteactions of

SAT

Road network

traffic flow












21 3 4 5

7 6 8 9 10

1211 13 14 15

1716 18 19 20

2221 23 24 25

Od

Dd

Currently empty SAT

Expected-empty SAT


Passed link

Remaining link


Origin of a trip













Time segments

Links

+

+

+

+

1ℎ

+ ge +

1ℎ+2Δℎ

1ℎ+3Δℎ

1ℎ+4Δℎ





+ ge + mℎ+4Δℎ

+ minus ge +


54













Demand

Waiting listNo









Tmin is found

E-E SATNo

YesC-S SAT

NoYes


YesE-S SAT

YesC-E SAT










21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(a)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(b)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(c)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(d)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(e)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(f)







119875( 119894119862119899) =119875119878119894119899119890119881119894119899

sum119895isin119862119899 119875119878119895119899119890119881119895119899(1)

119875119878119894119899 = sum119886isinΓ119894

( 119897119886119871 119894)1

sum119895isin119862119899 (119871 119894119871119895) 120575119886119895 (2)




Links

Time segments

1

+



Time segments

Links






+ minus ge +

+

+

+

+ 4












6

89 7

1610 1812 11

17

191514

2223

202113 24

1

3 4 5

23 km

2 km

4 km

14 km

41 km

2 km

14 km2 km















1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24Start Hour


0

1000

2000

3000

4000

5000

Dem

and

(Tri

ps)



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day


0 10 20 30 40 50 60 70 80Simulation day


085

09

095

1

105

11

115

12

125

13

135

Tim

e (m

in)

(c) Waiting time




10 20 30 40 50 60 70 800Simulation day

105

11

115

12

125

13

135

Tim

e (m

in)


0 10 20 30 40 50 60 70 80Simulation day


105

11

115

12

125

13

135

Tim

e (m

in)




03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day









Acknowledgments


References











































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia







Demand generation

SAT assignment

Enrouteactions of

SAT

Road network

traffic flow












21 3 4 5

7 6 8 9 10

1211 13 14 15

1716 18 19 20

2221 23 24 25

Od

Dd

Currently empty SAT

Expected-empty SAT


Passed link

Remaining link


Origin of a trip













Time segments

Links

+

+

+

+

1ℎ

+ ge +

1ℎ+2Δℎ

1ℎ+3Δℎ

1ℎ+4Δℎ





+ ge + mℎ+4Δℎ

+ minus ge +


54













Demand

Waiting listNo









Tmin is found

E-E SATNo

YesC-S SAT

NoYes


YesE-S SAT

YesC-E SAT










21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(a)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(b)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(c)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(d)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(e)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(f)







119875( 119894119862119899) =119875119878119894119899119890119881119894119899

sum119895isin119862119899 119875119878119895119899119890119881119895119899(1)

119875119878119894119899 = sum119886isinΓ119894

( 119897119886119871 119894)1

sum119895isin119862119899 (119871 119894119871119895) 120575119886119895 (2)




Links

Time segments

1

+



Time segments

Links






+ minus ge +

+

+

+

+ 4












6

89 7

1610 1812 11

17

191514

2223

202113 24

1

3 4 5

23 km

2 km

4 km

14 km

41 km

2 km

14 km2 km















1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24Start Hour


0

1000

2000

3000

4000

5000

Dem

and

(Tri

ps)



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day


0 10 20 30 40 50 60 70 80Simulation day


085

09

095

1

105

11

115

12

125

13

135

Tim

e (m

in)

(c) Waiting time




10 20 30 40 50 60 70 800Simulation day

105

11

115

12

125

13

135

Tim

e (m

in)


0 10 20 30 40 50 60 70 80Simulation day


105

11

115

12

125

13

135

Tim

e (m

in)




03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day









Acknowledgments


References











































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



21 3 4 5

7 6 8 9 10

1211 13 14 15

1716 18 19 20

2221 23 24 25

Od

Dd

Currently empty SAT

Expected-empty SAT


Passed link

Remaining link


Origin of a trip













Time segments

Links

+

+

+

+

1ℎ

+ ge +

1ℎ+2Δℎ

1ℎ+3Δℎ

1ℎ+4Δℎ





+ ge + mℎ+4Δℎ

+ minus ge +


54













Demand

Waiting listNo









Tmin is found

E-E SATNo

YesC-S SAT

NoYes


YesE-S SAT

YesC-E SAT










21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(a)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(b)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(c)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(d)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(e)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(f)







119875( 119894119862119899) =119875119878119894119899119890119881119894119899

sum119895isin119862119899 119875119878119895119899119890119881119895119899(1)

119875119878119894119899 = sum119886isinΓ119894

( 119897119886119871 119894)1

sum119895isin119862119899 (119871 119894119871119895) 120575119886119895 (2)




Links

Time segments

1

+



Time segments

Links






+ minus ge +

+

+

+

+ 4












6

89 7

1610 1812 11

17

191514

2223

202113 24

1

3 4 5

23 km

2 km

4 km

14 km

41 km

2 km

14 km2 km















1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24Start Hour


0

1000

2000

3000

4000

5000

Dem

and

(Tri

ps)



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day


0 10 20 30 40 50 60 70 80Simulation day


085

09

095

1

105

11

115

12

125

13

135

Tim

e (m

in)

(c) Waiting time




10 20 30 40 50 60 70 800Simulation day

105

11

115

12

125

13

135

Tim

e (m

in)


0 10 20 30 40 50 60 70 80Simulation day


105

11

115

12

125

13

135

Tim

e (m

in)




03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day









Acknowledgments


References











































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



Time segments

Links

+

+

+

+

1ℎ

+ ge +

1ℎ+2Δℎ

1ℎ+3Δℎ

1ℎ+4Δℎ





+ ge + mℎ+4Δℎ

+ minus ge +


54













Demand

Waiting listNo









Tmin is found

E-E SATNo

YesC-S SAT

NoYes


YesE-S SAT

YesC-E SAT










21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(a)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(b)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(c)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(d)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(e)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(f)







119875( 119894119862119899) =119875119878119894119899119890119881119894119899

sum119895isin119862119899 119875119878119895119899119890119881119895119899(1)

119875119878119894119899 = sum119886isinΓ119894

( 119897119886119871 119894)1

sum119895isin119862119899 (119871 119894119871119895) 120575119886119895 (2)




Links

Time segments

1

+



Time segments

Links






+ minus ge +

+

+

+

+ 4












6

89 7

1610 1812 11

17

191514

2223

202113 24

1

3 4 5

23 km

2 km

4 km

14 km

41 km

2 km

14 km2 km















1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24Start Hour


0

1000

2000

3000

4000

5000

Dem

and

(Tri

ps)



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day


0 10 20 30 40 50 60 70 80Simulation day


085

09

095

1

105

11

115

12

125

13

135

Tim

e (m

in)

(c) Waiting time




10 20 30 40 50 60 70 800Simulation day

105

11

115

12

125

13

135

Tim

e (m

in)


0 10 20 30 40 50 60 70 80Simulation day


105

11

115

12

125

13

135

Tim

e (m

in)




03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day









Acknowledgments


References











































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia




Demand

Waiting listNo









Tmin is found

E-E SATNo

YesC-S SAT

NoYes


YesE-S SAT

YesC-E SAT










21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(a)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(b)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(c)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(d)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(e)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(f)







119875( 119894119862119899) =119875119878119894119899119890119881119894119899

sum119895isin119862119899 119875119878119895119899119890119881119895119899(1)

119875119878119894119899 = sum119886isinΓ119894

( 119897119886119871 119894)1

sum119895isin119862119899 (119871 119894119871119895) 120575119886119895 (2)




Links

Time segments

1

+



Time segments

Links






+ minus ge +

+

+

+

+ 4












6

89 7

1610 1812 11

17

191514

2223

202113 24

1

3 4 5

23 km

2 km

4 km

14 km

41 km

2 km

14 km2 km















1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24Start Hour


0

1000

2000

3000

4000

5000

Dem

and

(Tri

ps)



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day


0 10 20 30 40 50 60 70 80Simulation day


085

09

095

1

105

11

115

12

125

13

135

Tim

e (m

in)

(c) Waiting time




10 20 30 40 50 60 70 800Simulation day

105

11

115

12

125

13

135

Tim

e (m

in)


0 10 20 30 40 50 60 70 80Simulation day


105

11

115

12

125

13

135

Tim

e (m

in)




03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day









Acknowledgments


References











































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(a)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(b)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(c)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(d)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(e)

21 3 4

6 5 7 8

109 11 12

1413 15 16

O1

O2

D1

D2

(f)







119875( 119894119862119899) =119875119878119894119899119890119881119894119899

sum119895isin119862119899 119875119878119895119899119890119881119895119899(1)

119875119878119894119899 = sum119886isinΓ119894

( 119897119886119871 119894)1

sum119895isin119862119899 (119871 119894119871119895) 120575119886119895 (2)




Links

Time segments

1

+



Time segments

Links






+ minus ge +

+

+

+

+ 4












6

89 7

1610 1812 11

17

191514

2223

202113 24

1

3 4 5

23 km

2 km

4 km

14 km

41 km

2 km

14 km2 km















1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24Start Hour


0

1000

2000

3000

4000

5000

Dem

and

(Tri

ps)



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day


0 10 20 30 40 50 60 70 80Simulation day


085

09

095

1

105

11

115

12

125

13

135

Tim

e (m

in)

(c) Waiting time




10 20 30 40 50 60 70 800Simulation day

105

11

115

12

125

13

135

Tim

e (m

in)


0 10 20 30 40 50 60 70 80Simulation day


105

11

115

12

125

13

135

Tim

e (m

in)




03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day









Acknowledgments


References











































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



Links

Time segments

1

+



Time segments

Links






+ minus ge +

+

+

+

+ 4












6

89 7

1610 1812 11

17

191514

2223

202113 24

1

3 4 5

23 km

2 km

4 km

14 km

41 km

2 km

14 km2 km















1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24Start Hour


0

1000

2000

3000

4000

5000

Dem

and

(Tri

ps)



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day


0 10 20 30 40 50 60 70 80Simulation day


085

09

095

1

105

11

115

12

125

13

135

Tim

e (m

in)

(c) Waiting time




10 20 30 40 50 60 70 800Simulation day

105

11

115

12

125

13

135

Tim

e (m

in)


0 10 20 30 40 50 60 70 80Simulation day


105

11

115

12

125

13

135

Tim

e (m

in)




03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day









Acknowledgments


References











































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



6

89 7

1610 1812 11

17

191514

2223

202113 24

1

3 4 5

23 km

2 km

4 km

14 km

41 km

2 km

14 km2 km















1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24Start Hour


0

1000

2000

3000

4000

5000

Dem

and

(Tri

ps)



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day


0 10 20 30 40 50 60 70 80Simulation day


085

09

095

1

105

11

115

12

125

13

135

Tim

e (m

in)

(c) Waiting time




10 20 30 40 50 60 70 800Simulation day

105

11

115

12

125

13

135

Tim

e (m

in)


0 10 20 30 40 50 60 70 80Simulation day


105

11

115

12

125

13

135

Tim

e (m

in)




03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day









Acknowledgments


References











































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24Start Hour


0

1000

2000

3000

4000

5000

Dem

and

(Tri

ps)



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



105

11

115

12

125

13

135

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day


0 10 20 30 40 50 60 70 80Simulation day


085

09

095

1

105

11

115

12

125

13

135

Tim

e (m

in)

(c) Waiting time




10 20 30 40 50 60 70 800Simulation day

105

11

115

12

125

13

135

Tim

e (m

in)


0 10 20 30 40 50 60 70 80Simulation day


105

11

115

12

125

13

135

Tim

e (m

in)




03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day









Acknowledgments


References











































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia




10 20 30 40 50 60 70 800Simulation day

105

11

115

12

125

13

135

Tim

e (m

in)


0 10 20 30 40 50 60 70 80Simulation day


105

11

115

12

125

13

135

Tim

e (m

in)




03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day



03

04

05

06

07

08

09

1

11

12

13

Tim

e (m

in)

10 20 30 40 50 60 70 800Simulation day









Acknowledgments


References











































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia



Acknowledgments


References











































RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia











RoboticsJournal of




VLSI Design



Shock and Vibration







Journal of



Volume 2018



Volume 2018


Journal of




SensorsJournal of



RotatingMachinery





Propagation






Hindawi


Advances in

Multimedia


shared autonomous taxi system and utilization of collected...

Documents