Enhanced Mobility With Connectivity and Automation: A Review of Shared Autonomous Vehicle Systems
11 Enhanced Mobility with Connectivity andAutomation: A Review of Shared AutonomousVehicle Systems
Liuhui Zhao,
Member, IEEE,
Andreas A. Malikopoulos,
Senior Member, IEEE
Abstract —Shared mobility can provide access to transportationon a custom basis without vehicle ownership. The advent of con-nected and automated vehicle technologies can further enhancethe potential benefits of shared mobility systems. Although theimplications of a system with shared autonomous vehicles havebeen investigated, the research reported in the literature hasexhibited contradictory outcomes. In this paper, we present asummary of the research efforts in shared autonomous vehiclesystems that have been reported in the literature to date anddiscuss potential future research directions.
Index Terms —Shared mobility, carsharing, connected andautomated vehicles
I. I
NTRODUCTION
A. Movitation
In a rapidly urbanizing world, we need to make fundamentaltransformations in how we use and access transportation.We are currently witnessing an increasing integration of ourenergy and transportation which, coupled with the humaninteractions, is giving rise to a new level of complexity [1]in emerging transportation systems such as connected andautomated vehicles (CAVs) and shared mobility. As we moveto increasingly complex emerging transportation systems, newcontrol approaches [2], [3] are needed to optimize their impacton the mobility system behavior.Shared mobility includes a variety of service models (e.g.,carsharing, ridesharing, bikesharing) to meet travel needs andmay result in a transformative impact on urban mobility [4]–[8] and landscape. As shared mobility services keep growing,there has been widespread speculation on their impact [7],[9], [10]. The advent of intelligent transportation systemsand information technologies has aimed at facilitating sharedmobility services (Fig. 1). In this context, impact analysis ofthe introduction of connected vehicles and automated vehicles(AVs) into existing shared mobility service is vital to identifythe opportunities and challenges related to shared autonomousmobility system. In this paper, we review the research reportedin the literature on carsharing enhanced by vehicle connectivityand automation technology, i.e., shared autonomous vehicle(SAV) system, and discuss potential implications in the envi-ronment and urban mobility.
This research was supported in part by ARPAE’s NEXTCAR program underthe award number de-ar0000796 and by the Delaware Energy Institute (DEI).The authors are with Department of Mechanical Engineering, University ofDelaware, Newark, DE 30332; e-mails: ([email protected], [email protected]).
B. Background
There are different types of carsharing service models,including round-trip carsharing, one-way station-based or free-floating carsharing, and peer-to-peer carsharing [5], [11]. Inthe past few years, short-term vehicle rental services providedby carsharing companies in major cities has attracted millionsof users, while the number is expected to grow significantly[9], [10], [12]. Generally, it is believed that carsharing haspositive impacts on energy use and greenhouse gas (GHG)emissions [12]–[16], particularly when low-polluting vehiclesare introduced into the transportation systems [17]. Althoughthere is evidence that the use of carsharing services leads toa decrease in vehicle ownership [12]–[14], location-specificvariations (e.g., urban form, level of transit service, availabilityof alternative modes, etc.) has an impact on vehicle milestraveled (VMT) and public transit ridership [9], [10], [12],[13], [18].The emerging CAV technologies offer intriguing opportu-nities to enhance urban mobility and traffic safety, and theintroduction of CAVs enables innovative often more responsiveand efficient options for traveling which may change the waypeople use mobility services [19], [20]. It is likely that thewide adoption of CAVs could also affect the usage of existinginfrastructure to better serve the ever-changing transportationnetwork [21]. While the benefits of CAV technologies ontraffic flow and safety [22]–[26], coordination in specific trafficscenarios [27]–[33], and energy improvement on vehicle levelFig. 1: A view of a city enhanced by connectivity andautomation. a r X i v : . [ m a t h . O C ] M a y [34]–[36] are well understood, potential deployment of theCAVs for the shared mobility service has raised a numberof key questions related to fleet sizing, operation strategiesand the implications on mobility, urban form, and environment[37]–[41].With the ongoing growth of shared mobility and increasinginterests in CAV fleet, the convergence of both emergingmobility service and technology is still evolving. Many majorautomakers and technology companies are launching SAVpilot projects in the US and around the world, e.g., Ford,Voyage, Waymo, Uber, and Lyft [42]. While there is cur-rently no large-scale deployment of SAV fleet, research effortshave studied impacts of the SAVs, including simulation-basedevaluation on environmental impact, cost-benefit, or demandanalysis, e.g., [43]–[57]. There has been much contention onthe potential influence of SAVs on travel behavior, urban land-scape, congestion, and environment [58]. Although it seemsthat the required fleet size as well as the parking spaces tomeet existing travel demand might drop significantly, multiplestudies have indicated that full automation is likely to inducetravel demand and attract new user groups, which may resultin a potential increase in energy consumption, e.g., [59], [60].Furthermore, there have been also concerns that SAVs mightattract considerable attention from public transit patrons ratherthan private car owners, with implications on escalating trafficcongestion, if not properly managed, e.g., [61], [62]. C. Scope of the Paper
In this paper, we review research efforts on the modelingand operations of the SAV system and try to identify potentialresearch gaps that require further investigation. In our review,we have excluded studies on the demand estimation and travelbehavior analysis of the SAVs. We applied the followingsearch strings and included the papers up to date containingany combination of the keywords in the title, abstract, orkeywords:1) shared autonomous (electric) vehicle(s), shared auto-mated vehicle(s), autonomous vehicle sharing;2) autonomous carsharing, driverless carsharing, self-driving carsharing;3) autonomous taxi, automated taxi, driverless taxi;4) automated demand responsive transport, autonomousmobility on demand, automated mobility on demand,autonomous mobility as a serviceAlthough the exploration of benefits of SAVs is still in earlystages, we note that there are many aspects in common withthe conventional carsharing system (with or without the optionof ridesharing). There are several review papers providing agood summary under the umbrella of shared mobility, e.g.,see [63]–[69]. Similar review efforts on the SAVs include thestudy by Hao and Yamamoto [70], who focused on the featuresand demand aspects of the SAV system through examining thecorresponding aspects of car sharing in AVs. The most recentwork conducted by Stocker and Shaheen [42] reviewed SAVpilots and legislation in the US, and discussed current andfuture development of the SAV system. Any such effort hasobvious limitations. Space constraints limit the description of each paper in details, and thus, discussions are included onlywhere they are important for understanding the fundamentalconcepts or explaining significant departures from previouswork.
D. Organization of the Paper
The structure of the paper is organized as follows. InSection II, we present an overview of the SAV system andmodeling approaches that have been commonly adopted. Wethen identify major design variables and system operatingparameters that are widely studied in the literature to date andsummarize the research efforts in Section III, including theproblems of fleet sizing, vehicle assignment and relocation,consideration of electric vehicles, and ridesharing. In SectionIV, we discuss different operation schemes of the SAVs ina mixed traffic environment that have been investigated inthe literature. Finally, we discuss research gaps and potentialfuture research directions in Section V.II. S
HARED AUTONOMOUS VEHICLE SYSTEM MODELING
SAVs provide carsharing with a way of seamlessly relo-cating vehicles to better match dynamic demand [46]. As thepilot programs of SAVs are beginning to accelerate aroundthe world, we observe an increasing interest in investigatingthe SAV system. In this section, we first introduce earlierwork on the feasibility of statewide implementation of SAVsand system performance analysis along with the cost-benefitanalysis. We then discuss two major directions in modelingand analysis of the SAV system: (1) the development ofanalytical models along with specific problems that includevehicle assignment and rebalancing, e.g., [71]–[74]; (2) thedevelopment of agent-based models to emphasize the under-standing of system performance and impact of the SAV systemunder different scenarios with a variety of parameters settings,e.g., [61], [75]–[79].
1) Feasibility analysis:
In an early work [43], Ford pro-posed a statewide SAV system in New Jersey with a grid-based network model. The author discussed different operationstrategies of a SAV system at different time periods. Forexample, in rush hours, the SAVs would function like apersonal rapid transit (PRT) system to satisfy travel demandand ease congestion, whereas during non-rush hours, the SAVscould be operated with more flexibility and provide door-to-door service. The area considered for the paper was modeledas gridded zones, where a fixed SAV station would be locatedat the center of each cell. Later, Brownell and Kornhauser [80]described in detail two distinct SAV network models, i.e., PRTand the smart paratransit (SPT), and discussed the feasibilityof a statewide SAV network in New Jersey. In the PRTnetwork, fixed stations of the SAV system are established andpassengers need to walk to their closest stations. Ridesharingis considered only if two passengers share the same origin-destination pair and arrive at the station within a predefinedtime window. The idea behind the SPT system is that tripswith close origins and/or destination will be served by onesingle vehicle. The vehicle moves around within the origincell to pick up multiple passengers before traveling to the destination cell. Along the ways, the vehicle may stop atone, or more, locations to pick up or drop off passengers.In a SPT system with AVs, since the vehicle take the placeof the individual for accessing service, the distance betweennodes in the transit grid could be increased. Burns et al. [44]conducted a cost-benefit analysis of a SAV system wherethe entire trip demand is satisfied by SAVs. To estimatethe performance of a SAV system and compare with othersystems (e.g., personal vehicle), the authors developed ananalytical model with spatial queueing approach based onsimplifying assumptions (e.g., uniformly distributed originsand destinations, constant trip request rate, etc). The resultsfrom three case studies showed that a SAV system is capableof providing better mobility experience at a significantly lowercost, in addition to its environmental and safety benefits.
2) Analytical modeling:
Several research efforts reportedin the literature have treated a SAV system as a spatialqueueing system where passengers arrive at each station, pickup the vehicles – if parked at the station – and wait orleave the system, if no vehicle is available (Fig. 2). Afterdropping off passengers at their destinations, vehicles eitherstart the next service, or park, or relocate themselves to otherstations, e.g., [71]–[74], [81], [82]. For instance, Zhang et al.[38] described a SAV network as a spatial queueing systemwhere transportation requests queue up and are served by theSAVs in the network. The authors presented two models forSAV systems: (1) in the first model, the authors considereda distributed approach, where the objective is to design arouting policy that minimizes the average steady-state timedelay between the generation of an origin-destination pair andthe time the trip is completed; and (2) in the second model,the authors considered a lumped approach – customers areassumed to arrive at a set of stations in the network, whereeach customer picks up a vehicle, if available, or leaves thesystem, if no vehicle is parked at the station.Fig. 2: Shared autonomous vehicles in a queueing system.
3) Agent-based modeling:
To address the questions on theimpact of SAVs on transportation mobility and investigateperformance of the SAV system under various scenarios,several research efforts have also focused on developing agent-based models to evaluate the transportation network withpresence of SAVs [61], [75], [76], [83]. With the advantageof modeling each individual passenger/vehicle as an agent following simple rules, complex behavior [84], [85] at amacroscopic level emerges, which provides an approximationof travel behavior in the transportation systems [75]. Marczuket al. [86] and Azevedo et al. [87] proposed an extension tothe agent-based demand and supply model (SimMobility) forthe design and evaluation of the SAV system in a multi-levelsimulation framework, and explored the effects of fleet sizeand station location for both station-based and free-floatingSAV systems. Boesch and Ciari [75] discussed the advantagesof MATSim (an activity-based agent-based simulation model)with the presence of SAVs and its potential applicationson investigating related problems, such as “the potential ofSAVs complementing or competing with other transporta-tion modes,” “appropriate fleet size in different transportationsystems,” and “the demand distribution with respect to theresponse of different fleet sizes.”Focusing on the potential impact of SAV system on urbanparking demand, Zhang et al. [77], [78] investigated differentsystem operation strategies under low penetration of SAVswith an agent-based simulation model. Ridesharing and trav-eler’s acceptance of sharing rides were also explored in thepaper. The results showed a significant parking demand reduc-tion with the SAV system – enabling ridesharing and addingvehicle cruising options would further reduce parking demand.Kondor et al. [88] developed an agent-based simulation modelto estimate parking demand savings with shared vehicles andSAVs for home-work commuting. Similar conclusions weredrawn from this study that up to 50% reduction in parkingneeds could be achieved at the expense of less than 2%increase in VMT. Jager et al. [83] developed an agent-basedframework for a shared autonomous electric vehicle (SAEV)system that reflect the system behavior on an operationallevel. Although the system has a central dispatcher, the ve-hicles compete for customers and make their own decisionsfor routing and charging. Simulation results confirmed thefeasibility of operating a SAV fleet with both high servicelevels and vehicle utilization, however, environmental benefitscan only be expected when using renewable energy sourcesand enabling ride sharing features.III. S
HARED AUTONOMOUS VEHICLE SYSTEM DESIGNVARIABLES
Similar to conventional carsharing service, not only theoperations of a SAV is significantly affected by the assignmentand rebalancing strategies over a fleet of SAVs, mobility andenvironment, but also the urban landscape can be considerablyimpacted by the implementation strategies of a SAV system.Naturally, the problems of fleet sizing, vehicle-trip assignment,and rebalancing in a network of SAVs are the major subjectsin enhancing our understanding of a SAV system, with theoptions of ridesharing and usage of electric vehicles thathave attracted considerable attention recently. The majorityof the literature to date has concentrated on how the SAVsystem tackles one or more of the aforementioned problems,and has aimed at enhancing our understanding about theperformance and potential benefits of the network with a fleetof SAVs. In the following subsections, we provide a summary of SAV system modeling and discuss key topics that have beeninvestigated in previous studies regarding the SAV system.
A. Fleet Sizing of a Shared Autonomous Vehicles System
Fleet size is the major determinant of the operating costof the SAV system. General considerations in determining thefleet size include system access, directness, sharing, and pas-senger waiting time [89], [90]. In what follows, we summarizedifferent approaches in addressing fleet sizing problems in aSAV system.Fagnant et al. [91] simulated a SAV system in Austin areawith a grid-based network model following a similar modelingframework presented in [43]. In this work, a fleet of SAVsis generated in the network to ensure that passenger waitingtimes are within predefined bounds. A heuristic strategy isimplemented to relocate vehicles such that the stock of SAVsamong cells is balanced. A replacement rate of 1 SAV per9.3 conventional vehicles was identified as appropriate forthe area considered. The authors concluded that even withan excess VMT, emissions and environmental outcomes forthe SAVs are still advantageous compared to those for theaverage US vehicle fleet. In the modeling framework for theSAV system developed by Winter et al. [89], the minimumfleet size and the optimal fleet size that yield the minimumsystem costs are determined through an iterative procedure,where the core is a simulation tool that is applied for assigningvehicles to passenger requests. Several scenarios are conductedto analyze the influence of different design parameters (e.g.,vehicle capacity, operational parameters, demand level) onsystem performance.Vazifeh et al. [92] investigated the minimum fleet sizeproblem of a SAV system with a network-based model. Tripsbased on known demand and link travel times were takenas input to construct the vehicle shareability network underthe constraint of maximum trip connection time. With fullyknowledge of daily trip demand, the authors found that 40%taxis in New York City can be reduced without incurring delayto passengers, under the constraint of 15-minute maximum tripconnection time. Relaxing the assumption of complete demandinformation, the authors concluded that if trip requests werecollected at 1-minute interval, the system could be operatedwith a 30% fleet reduction at a relative high level of service(i.e., above 90% served trips within a 6-min delay).Spieser et al. [73] addressed two major fleet sizing prob-lems: (1) the minimum number of vehicles needed to stabilizethe workload of a SAV system and (2) the number of vehiclesneeded to ensure a desired level of service provided to thecustomers. In their paper, the SAV system is modeled as aqueueing network where each region is mapped into single-server node, and each route between each pair of regions ismapped into infinite-server nodes. The vehicle rebalancingprocess is modeled as an arrival process of “virtual passen-gers.” Conducting a case study in Singapore, the paper showedthat a SAV can meet the personal mobility needs of the entirepopulation with a fleet size about one third of the total numberof passenger vehicles currently in operation.Masoud and Jayakrishnan [93] discussed a different imple-mentation strategy of the SAV system, with households form clusters (i.e., neighborhoods). Each neighborhood share theownership and ridership of a set of autonomous vehicles thatserve as rental cars during their idling times. The authorsfocused on the optimization of the fleet size in a neighborhoodand the number of rental requests for the vehicles during theiridling times. Two optimization models were developed. Thefirst model addressed the neighborhood clusters and aimed atminimizing the total number of the vehicles by consideringessential trips to be satisfied for all the households in aneighborhood. The second model optimized the total numberof rental requests so as to maximize extra income from idlingvehicles, considering time window constraints of the owners’essential trips.Most of previous work has emphasized on searching forthe minimum fleet size of SAVs that could provide serviceon the existing demand at a desired level, when replacingthe existing conventional vehicle service by SAVs. We havenoticed promising results from multiple papers indicating thata high replacement rate of conventional vehicles is feasibleto satisfy the same level of demand. However, there is stillsome additional work missing to assess holistically the impactof urban mobility due to potentially changing travel behaviorand demand as a result of the introduction of AV in the mixedtraffic environment.
B. Vehicle Assignment in a Shared Autonomous Vehicle System
Although there is a rich body in the literature in dynamicassignment problems with various applications on taxi, para-transit, trucking services, etc, that require real-time vehicleassignment to dynamic service requests (e.g., see [94]–[97]for more details), most papers reported in the literature to datehave focused on investigating SAV system performance withsimplified vehicle assignment strategies (usually rule-based).In what follows, we present a general formulation of thevehicle assignment problem in a SAV system. Let i ∈ M be atrip request, j ∈ N be the index of a vehicle, and x ij equal to1 if and only if trip i is assigned to vehicle j , where M ⊂ N isthe set of trip requests and N ⊂ N is the set of vehicles. Thegeneral vehicle-traveler, or vehicle-trip, assignment problemto minimize the objective function J a of trip assignment costis as follows [38], [98]: min J a = (cid:88) i (cid:88) j c ij x ij , (1)subject to (cid:88) j x ij = 1 , i ∈ M , (2) x ij ∈ { , } , ∀ i ∈ M , j ∈ N , (3)where c ij is the cost of assigning trip request i to vehicle j ,which could be represented by trip travel distance, travel time,or monetary cost. The trip assignment cost in (1) is evaluated atevery trip assignment time step with dynamic service requests.The constraint (2) ensures that each traveler is assigned to onlyone vehicle.When assigning travelers to the nearest idling AVs, sev-eral research efforts have considered a first-come-first-served strategy, which is a heuristic approach to minimize passengerwaiting time [44], [78], [90], [91], [99]. In a paper by Fagnantand Kockelman [46], the SAV service area is divided intosmall zones, where trips are randomly generated. Every fiveminutes, passengers will be randomly ordered and assigned tothe nearest available SAV in the same zone, up to a maximumvehicle arrival time. If such assignment fails, those passengerswill be held until next assignment. Hyland and Mahmassani[100] investigated the underlying stochastic vehicle assignmentproblem for the SAV system with no shared rides. With theassumption that the fleet operator has no information of thespatial-temporal demand distribution, the authors compareddifferent SAV assignment policies as the solution approachesto the local optimization problem at each time step. Two of theapplied strategies were first-come-first-served, and the otherstrategies minimized traveler waiting times (under differentvehicle-traveler assignment constraints).Hanna et al. [101] examined different methods for assigningvehicles in a SAV system: (1) a decentralized greedy matchingwhere users are assigned to their nearest vehicles in a randomorder, (2) a centralized greedy matching approach ensuringthat each vehicle is matched with its closest user, (3) theHungarian minimum cost matching algorithm that minimizespassenger waiting time and unoccupied distance traveled, and(4) a minimal makespan matching algorithm which minimizesthe longest distance that any vehicle must travel to a passenger.The authors showed that compared to greedy approaches, thelatter two algorithms improved system performance throughreducing unoccupied travel distance, passenger waiting time,and waiting time variation. C. Vehicle Rebalancing of a Shared Autonomous Vehicle Sys-tem
The SAV system shares similar characteristics with thecarsharing system consisting of conventional vehicles [46].In terms of unbalanced demand distribution, both systemsface the same problem of vehicle rebalancing. Two majorrebalancing strategies have been investigated in the litera-ture of carsharing with conventional vehicles including (1)operator-based vehicle relocation and (2) user-based vehiclerelocation, which could potentially be adapted in addressingthe same problem in the SAV system, see [63], [102]–[105].However, the relocation of SAVs still have differences withthat of conventional sharing vehicles, in that the SAVs arefully compliant and always cooperative [106]. Thus, due tothe inherent capabilities of self-driving and self-rebalancingof a SAV system, research efforts have focused more onthe problem with a centralized operator that has dispatchingcontrol over the entire SAV network, which may yield a systemoptimum solution for the entire system.We provide a general formulation to illustrate vehicle rebal-ancing problem for a SAV system. Let r y be the number ofidling vehicles in zone/station y ∈ Z and r yz be the numberof rebalancing vehicles from zone/station y to zone/station z ∈ Z , where Z ⊂ N is the total number of zones/stations inthe network. Generally, the objective function J r is the totalcost induced by vehicle rebalancing [71], [107], [108]: min J r = (cid:88) y (cid:88) z c yz r yz , (4)subject to (cid:88) z r yz = r y , ∀ y, z ∈ Z , (5) r yz ∈ N , ∀ y, z ∈ Z , (6)where c yz is the cost of moving vehicles from zone/station y to zone/station z , which could be represented by triptravel distance, travel time, or monetary cost. In a systemwith dynamic trip requests, (4) will be evaluated at everyrebalancing time step and (5) defines the total rebalancingvehicles from zone/station y should equal the number of idlingvehicles in the zone.Targeting at the problem of unbalancing demand and supply,Pavone et al. [109] addressed the vehicle relocation problemfor a mobility-on-demand system, optimizing the rebalancingassignment that minimizes the number of vehicles to bemoved. Using a fluid model of the system, they showed thatthe optimal rebalancing policy can be found as the solutionto a linear program, under which every station reaches anequilibrium where there are excess vehicles and no waitingcustomers. Based on this study, Zhang and Pavone [72]presented a queueing-theoretical approach and provided thesolution to an offline optimal rebalancing problem. Later, Wenet al. [107] extended the research by incorporating door-to-door service and ridesharing option in a free-floating SAVsystem. From the fleet operator’s perspective, Spieser et al.[108] investigated the vehicle rebalancing problem in a SAVsystem by quantifying the operation cost as a function of fleetsize, demand loss and utilization rate, and analyzed the impactof fleet size on demand loss, vehicle utilization rate, andvehicle rebalancing miles traveled. H¨orl et al. [110] evaluatedperformance of four heuristic and optimal rebalancing policiesfor a SAV system in an agent-based simulation environment,and suggested that the utilization of intelligent demand fore-casts and rebalancing algorithms would be crucial for a SAVsystem to be competitive with private vehicles.Through simulation based evaluation, recent work focusedon the impact of vehicle rebalancing strategies in a SAVsystem. Zhu and Kornhauser [111] investigated the rebalancingstrategies for the SAV system in New Jersey and their effectson the fleet size and level of service provided in scenarioswhere all non-walking travel demand is served by SAVs.Shared trips are served by vehicles of different capacities (i.e.,3, 6, 15, and 50 passengers). Two rebalancing strategies aredeveloped based on known demand. In the first approach,vehicles are moved at the end of the day to make sure thatthere are enough vehicles at each station that satisfy thedemand at the beginning of the day. In the second approach,vehicles are relocated as needed to fill in any station withoutenough vehicles. The authors also evaluated the performanceof the statewide SAV system with varying fleet sizes, in termsof passenger waiting time and rebalancing trip lengths. Theresults showed that one SAV could possibly replace more thansix traditional vehicles while the demand could still be wellserved. Fagnant and Kockelman [46] investigated the operation ofSAVs through an agent-based model and focused on the im-plications of travel and environmental impacts of SAVs undera mixed traffic condition. Addressing the imbalanced demandpatterns, the authors proposed several relocation strategies tobalance vehicle supply and reduce future traveler wait times:(1) relocating vehicles based on expected demand and (2)relocating vehicles to balance stock based on predicted supply.Marczuk et al. [112] developed a simulation framework forrebalancing a one-way SAVs system in SimMobility environ-ment. The proposed fleet management center is responsiblefor passenger-to-vehicle assignment, vehicle routing and re-balancing. Three vehicle relocation strategies were proposedfor the system: (1) no rebalancing as the baseline scenario,(2) offline rebalancing that minimizes the number of rebal-ancing trips, and (3) online rebalancing that minimizes thetotal time/effort spent for rebalancing per rebalancing interval.Winter et al. [113] analyzed the impacts of different relocationstrategies of a SAV system in a simulated generic grid network.Five vehicle relocation strategies were tested, including re-maining idle, random shuffling, returning to original location,moving based on demand anticipation, and moving to balancevehicle stock over the network. In the simulation framework,the fleet size of the SAV system is given as an input, and vehi-cles are dispatched through a rule-based strategy. Performancemeasures such as average passenger utility, average waitingtime, and the ratio of vehicle driving time were examined.The simulation showed that remaining idle strategy wouldbe the most efficient in terms of passenger waiting time, yetthe worst performer considering link occupancy and parkingturnover rates. In contrast, strategies aiming at distributingvehicles yielded higher parking turnover rates but showedlower service efficiency. In light of these results, the authorsextended the study by imposing the constraints of limitedparking facilities in the evaluation of the above five heuristicrelocation strategies for idle SAVs, and examined the potentialimpact of SAVs on urban traffic in terms of congestion,parking consumption and mode shift [114].As discussed in the above papers, e.g., [111]–[114], de-pending on the objectives and targeting performance measures,the rebalancing strategy to be applied in a SAV system maybe different. The operation of a fleet of SAV is considerablyaffected by the applied relocation strategy or a combination ofstrategies, considering the inter-dependencies among parkingdemand, traffic condition, and user mode choice. Althoughcurrent research efforts emphasize rebalancing strategies in anisolated SAV system, the externalities should be analyzed inmore depth to enhance the understanding of traffic dynamicswith the implementation of SAV service.
D. The Usage of Electric Vehicles in a Shared AutonomousVehicle System
A significant amount of work has focused on the use ofelectric vehicles in a SAV system to achieve larger energyand emission savings for a greener transportation system [61],[76], [90]. Considering the range of electric vehicles, there isa number of constraints in a SAEV system. For instance, a vehicle may need to visit a charging station after droppingoff passengers. There may be instances that vehicles have toturn down trip requests and drive to charging stations instead,resulting in different vehicle-trip assignment strategies [115]–[117].Based on the work in [38], Zhang et al. [106] presented amodel predictive control (MPC) approach to optimize vehiclescheduling and routing in a SAEV system, considering vehiclecharging constraints. Compared to other control algorithms ofa SAV system (i.e., nearest-neighbor dispatch, collaborativedispatch, Markov redistribution, real-time rebalancing), theauthors concluded with a case study in New York City that theMPC algorithms outperformed the other strategies in terms ofaverage customer waiting times.Chen et al. [61], [76] addressed the operations of a SAEVswith an agent-based model based on the work reported in [46]and [91]. The emphasis of this research is the performanceanalysis of a fleet of SAEVs under various vehicle range andcharging infrastructure scenarios. The authors also exploredthe pricing schemes of a SAEV system when competingagainst other modes (i.e., private human-driven vehicles andcity bus service), and found that with higher SAEV penetrationrate, the private vehicle replacement rate by the SAEVsincreases, leading to improved system performance. Similarly,the study by Bauer et al. [118] predicted battery range andcharging infrastructure requirements of a fleet of SAEVsoperating on Manhattan island with an agent-based model.The authors also conducted sensitivity analysis of the costand the environmental impact of providing SAEV service witha wide range of changes in cost components (e.g., batterytype, vehicle type, etc.). The study indicated that instead ofbattery range, the major challenge to introducing SAEVs maybe building sufficient charging infrastructure.Kang et al. [115] developed a framework for a SAEV systemthat consists of demand forecasting, fleet assignment, electricvehicle designing, and charging station locating modules.The fleet assignment module determines the optimal vehicleassignment and charging schedules, and the charging stationlocating module decides the optimal charging station locations.The system-level objective is to maximize service profit for theoperator, through optimizing decision variables including fleetsize, number of charging stations, electric powertrain design,membership fee, and vehicle rental fee. The locations of charg-ing stations are selected with a p-median model from a poolof predetermined candidates. A comparison between a SAVsystem and a SAEV system was conducted in terms of costand benefit under different scenarios (e.g., varying gas pricesand charging station installation costs), showing that a SAEVsystem would be more profitable for most of the scenarios.Although both systems are marketable, the optimized SAEVsrequired longer waiting times than optimized SAVs due to theconstraints of vehicle range and charging issues.Iacobucci et al. [119] developed a simulation model toevaluate a SAEV system interacting with passengers andcharging at designated stations based on a heuristic chargingstrategy. The potential utilization of the SAEV system as anoperating reserve provider and its performance in response togrid operator requests were evaluated. The authors concluded that the proposed system could reduce the required fleet sizeas compared to private vehicles while providing a comparablelevel of transportation service with low break-even prices.Later, based on the work presented in [106], the authors devel-oped a framework for the optimization of charging schedulingand vehicle routing and relocation for a fleet of SAEVs [120].The proposed framework consists of two layers of optimizationmodel: over longer time scales, the charging scheduling opti-mization minimizes waiting times and electricity costs, whileover shorter time scales, vehicle routing and relocation areoptimized under charging constraints. The authors reportedthat a substantial reduction in charging costs was yieldedfrom the proposed framework without significantly affectingpassenger waiting times, as well as the potential of SAEVs tooffer energy storage to the grid and avoid grid congestion.In summary, the introduction of electric vehicles in the SAVsystem offers a large potential to further enhance environ-mental benefits. However, constraints such as vehicle rangeand charging facility locations add more dynamics into thesystem, and multiple studies suggested that the infrastructureand charging scheduling are the key influencing factors ofsystem performance of a fleet of SAEVs. Considerably workhas focused on the performance analysis of SAEV system ascompared to the SAV system, through evaluating the impactof vehicle range, charging infrastructure, as well as electric-ity costs [61], [76], [118]. Considering charging constraints,several research efforts have also emphasized on re-examiningvehicle routing and relocation strategies as well as optimizingcharging locations [106], [115], [121]. Recently, the optionof vehicle-to-grid as well as the integrated planning of powergrid and shared mobility service has also attracted considerableattention [119], [120], to improve the perception of SAEVsand ensure sustainable commutes within the notion of smartcities [122].
E. The Option of ridesharing in a Shared Autonomous VehicleSystem
The problems of ridesharing and carsharing are usuallydecoupled in the existing literature [123]. Recently researchefforts started exploring the option of ridesharing in a SAVsystem, e.g., [124]–[127]. By allowing ridesharing, the fleetsize may be further reduced to provide a desired level ofservice to the passengers, although the total VMT probablymight increase [128], [129]. There are generally two typesof ridesharing as illustrated in Fig. 3: (a) trip combiningneighboring origins and destinations (Fig. 3a) and (b) tripchaining based on trip temporal and spatial characteristics(Fig. 3b). We consider here ridesharing as the option of servingmultiple passengers in a single vehicle trip, or trip chain, inthe SAV system, and emphasize the impact of opening upridesharing options in the SAV service, without detailing theoperation modes and strategies for ridesharing. Consideringdifferent system objectives (e.g., minimizing total vehiclemiles traveled, minimizing total travel time, or maximizingserved trips) and various system constraints (e.g., time windowand seat constraints), there has been work on the SAV systemwith the option of ridesharing and the evaluation of differentridesharing strategies against network performance. (a)(b)
Fig. 3: ridesharing in the shared autonomous vehicle system:a) trip combination; b) trip chaining.Levin et al. [124] analyzed the possibility of ridesharing ina SAV system where passengers could select the first arrivedvehicle regardless of occupancy. The authors found that SAVswith the choice of ridesharing may cause more congestion dueto additional miles traveled for detouring. Zhang et al. [78],[142] applied an agent-based model to evaluate the perfor-mance and potential benefits of a SAV system with dynamicridesharing. In a grid-based simulation network, a centralizedoperator monitors real-time trip requests and SAV status aswell as manages trip assignment for the SAV system, whereridesharing option is evaluated against passenger’s willingnessand travel cost. Their work suggested that dynamic ridesharingin a SAV system could potentially lead to reduced vehicleownership, parking demand, and emissions.Hyland and Mahmassani [125] compared the performanceof a SAV system with and without ridesharing option in termsof the ability to handle demand surges. In this paper, the math-ematical formulations of the vehicle assignment with/withoutridesharing were presented and the solutions were derived witha rolling-horizon approach. The simulation results indicatedthat the SAV with ridesharing service improved system per-formance in response to demand surges.Based on the vehicle rebalancing strategies tested in [46],Fagnant and Kockelman [126] considered the option of dy-namic ridesharing in a simulated SAV system. With the casestudy of a 24-mile by 12-mile region in Austin, the authorsconcluded that dynamic ridesharing in a SAV system was ableto limit excess VMT from the SAV system, reduce passengerwaiting times (under the constraint that ridesharing should notincrease travel time of current passengers by more than 40%),and yield an enhanced level of service.Farhan and Chen [141] discussed the impacts of ridesharingon the operational efficiency of SAEVs with a discrete-timesimulation model. Both the fleet size and number of chargingstations are determined during simulation. In their research,
TABLE I: Approaches in Shared Autonomous Vehicle System Modeling.
Approach Topic ReferenceOptimization Fleet sizing [38], [73], [92], [93], [98], [115], [130]Vehicle routing / trip assignment [81], [98], [106], [120], [123], [125], [127], [131]–[136]Vehicle rebalancing / relocation [72], [81], [107], [108], [130]Other considerations [115], [121], [134]Simulation Evaluation Fleet sizing [43], [44], [47], [48], [80], [86], [89], [90], [128]Vehicle routing / trip assignment [83], [87], [99], [128], [137], [138]Vehicle rebalancing / relocation [46], [48], [77], [91], [104], [110], [112]–[114], [128], [137], [139], [140]Ridesharing [77], [124], [126], [129], [141]–[147]Pricing scheme [57], [61], [145], [148], [149]Transit integration / mode choice [61], [87], [89], [138], [140], [143], [146], [148]–[154]Electric vehicles [57], [76], [83], [104], [116]–[119], [140], [141] the travelers are grouped into clusters based on spatial criteria,and the ride-share matching problem is formulated as a vehiclerouting problem minimizing system-wide vehicle miles trav-eled under time window constraint. The results indicated thatallowing a second passenger in ridesharing yielded marginalbenefit of fleet size and charging station reduction. Althoughmore passengers in shared trips reduced the required fleetsize and number of charge stations, passenger waiting timesincreased due to ridesharing (i.e., reduced level of service).IV. S
HARED A UTONOMOUS V EHICLE S YSTEM O PERATION
Although the majority of the literature is focused on ex-amining the feasibility and performance of the SAV serviceas an isolated system, there is an increasing interest towardsthe investigation of more realistic operational scenarios relatedto the SAVs. Recent research efforts have also focused onanswering questions such as: “How will the SAV systemperform in a mixed traffic environment?” “What will be themobility impact of integrating the SAVs with other modes oftransport?” In this section, we focus on different operationalaspects of a SAV system, and summarize the studies thatconsider realistic and mixed traffic conditions.
A. Operation in a Realistic Traffic Environment
The majority of the aforementioned work has addressed theSAV system with full SAV penetration or without consideringbackground traffic. There are only a few papers focusing onthe congestion impact of SAVs, e.g., [81], [124], [132], [155],[156]. For example, to investigate mobility impacts of SAVs,Levin et al. [124] presented a general event-based frameworkfor simulating the operations of a SAV system with existingtraffic models. Considering 100% penetration of SAVs, theauthors found that under certain scenarios (e.g., with the optionof dynamic ridesharing), a smaller fleet of SAVs performedbetter than a larger fleet due to lower congestion in the net-work. Maciejewski and Bischoff [156] evaluated the impacts ofa city-wide introduction of SAVs on traffic congestion throughan agent-based simulation model, focusing on the analysisof traffic congestion under different SAV penetration rates.With an assumption of increased road capacity due to AVoperations, their work showed that despite increased trafficvolume, a fleet of SAV could have a positive effect on trafficat a penetration rate as low as 20%. Levin [131] developed a linear programming formulationfor vehicle routing problem in the SAV system, where trafficflow was modeled through the link transmission model. Theresults showed that asymmetric demand (e.g., demand duringpeak periods) could lead to significantly rebalancing trips andgreater congestion than uniformly distributed demand pattern.Since more vehicles might cause additional congestion onroadway network, it is important for the SAV system toplan for different traffic patterns. Liang et al. [132] proposedan integer programming model to define the routing of theSAVs based on profit maximization function, where traveltimes on the links varied with the flow of SAVs (withoutany background traffic). Later in [133], the authors appliedthe algorithm for trip assignment and dynamic routing inthe city of Delft, the Netherlands with a rolling horizonscheme. Assuming that the operator of a SAV fleet has thechoice of accepting or rejecting trip request according to profitmaximization function, this analysis showed that taking intoaccount the impact of dynamic travel time led to differentresults of satisfied trips and VMT, and ultimately affectedoverall operator profit and network congestion level.Rossi et al. [81] studied the routing and rebalancing problemof SAVs in congested transportation networks, where a SAVsystem is modeled in a network flow framework such thatvehicles are represented as flows in a road network. Theobjective of the routing problem is to minimize the weightedsum of passenger trip travel times and vehicle rebalancingtravel times considering network capacity. The objective of therebalancing problem is to optimize rebalancing paths such thattraffic congestion is minimized. Through numerical studies onreal-world traffic data, the authors showed that the proposedreal-time routing and rebalancing algorithm yielded lowercustomer waiting time by avoiding excess congestion on theroad, compared to point-to-point rebalancing algorithms whereno underlying road network is assumed.Through an agent-based model, Fagnant and Kockelman[46] investigated the operation of SAVs and focused on theimplications of travel and environmental impacts of SAVsunder a mixed traffic condition. Instead of 100% penetrationof SAVs, the authors considered the transportation systemwith a small market share of SAVs (i.e., around 3.5%). Thesimulation results under different scenarios (e.g., varying tripgeneration rates, network congestion levels, SAV fleet size,etc) indicated that each SAV can substitute around eleven conventional vehicles at the cost of 10% more VMT, and theoverall emissions savings are expected to be sizable for mostemission species.
B. Operating in a Multi-Modal Environment
Based on the discussion in the previous sections, it seemsclear that SAVs, compared to personal owned human-drivenvehicles, have significant advantages for individuals as well asfor the transportation system in terms of mobility, safety, andenergy savings (especially with SAEVs), e.g., [46]–[48], [76],[83], [86]. A combination of SAVs with other transportationmodes such as public transportation, however, might imposedifferent conclusions [9], [10], [12], [18]. Although SAVscould be utilized in the way to facilitate the first and last miletransport [157] and promote the use of public transportationsystem (e.g., [154], [158]), SAVs may also divert passengersaway from transit systems due to their capability of providingdoor-to-door services (e.g., [154], [159]).
1) Shared autonomous vehicles as a complement of publictransit:
Early research efforts have explored the performanceof integrating the SAV system with transit systems. Forexample, based on the same network in New Jersey as in [43],Zachariah et al. [143] simulated a system of SAVs where thetrain network is preserved and treated as an integral part ofthe system. Using SAVs as a complementary service of a trainsystem, Liang et al. [134] presented an optimization model todefine the service area of a SAV system for first/last miletransport that maximizes the profit of the SAV operator. Laterin [135], the authors designed a SAV system providing shuttleservice between a major train station and city area, consideringthe competition between SAVs or other modes (e.g., biking orwalking), as well as the impact of traffic congestion on modesplit. With the objective of minimizing total travel time, theauthors developed an optimization model to decide the bestfleet size and price rate for the SAV system.Shen et al. [146], [151] explored the feasibility of integrat-ing SAVs in the public transportation system to improve thefirst/last mile connectivity. With a simplified simulation modelwithout considering traffic congestion where the demand forthe SAV system was assumed to be 10% of the original busdemand, the study showed that by enabling ridesharing, theintegrated service was able to reduce average passenger traveltime and ease traffic through less occupancy of road resources.Scheltes and de Almeida Correia [140] studied the SAEVsystem providing last-mile service for a train line. In thesimulation model, vehicle assignment in response to travelerrequest followed a first-come-first-served model. Meanwhile,the scenarios of short-term pre-booking, vehicle relocating,and opportunity charging were also explored. The resultsshowed that compared to bicycle and walking as last miletransportation modes, the SAEV system was able to reduceaverage passenger travel time and waiting time, especiallywhen pre-booking option was enabled.Wen et al. [153] proposed a systematic approach to de-sign and simulate an integrated system of SAVs and publictransit. The authors emphasized that the SAV operation isdesigned to be transit-oriented with the purpose of supporting existing public transit service. In an agent-based simulationplatform, the interaction between service operator and travelersis modeled with a set of system dynamics equations, suchthat the decisions of both parties could be captured in thesystem. The authors suggested that encouraging ridesharing,allowing in-advance requests, and combining fare with transitwould be useful to enable service integration and promotesustainable travel. Pinto et al. [138] proposed a simulationframework integrating travel mode choice model and dynamictransit assignment model to assess the impacts of a suburbanfirst-mile SAV system on transit demand. Similarly, Martinezand Viegas [150] presented an agent-based model to evaluatethe impact of the SAVs in the city of Lisbon, Portugal. Intheir simulation model, current travel demand is served bytwo types of AVs that compete with each other, i.e., a SAVproviding door-to-door service with the choice of ridesharingand an autonomous minibus that replaces current bus servicewithout any transfers for users. The simulation results revealedpositive mobility impact of SAVs especially when introducingthe autonomous minibus into the network.
2) Shared autonomous vehicles as a competitor of publictransit:
Liu et al. [149] simulated transportation patterns inAustin network with a system of SAV from a mode-choiceperspective. A user-equilibrium based dynamic traffic assign-ment model was applied in simulation environment. The studyfocused on travelers’ mode choices with the presence of SAVs.In a mixed traffic environment, where private human-drivenvehicles, public transit, and SAVs coexist, the study analyzedthe impacts of the SAV system on energy consumption andemissions under different SAV penetration rates and SAVrental fees. Based on the sensitivity analysis of rental fees, theauthors found that if the SAV fare rate is low enough, SAVusers might travel more than private vehicle users. Therefore,although the use of AVs is expected to result in energysavings and emission reduction, the extra VMT by SAVs couldcompromise such environmental benefits. The mode choiceresults indicated that, for travelers who do not own a privatevehicle, SAVs are preferable for short-distance trips comparedto public transit – demand shifting from public transit would bea concern once the SAVs become available in the study area.H¨orl [148] conducted a similar study and investigated the SAVservice in a multi-modal traffic simulation environment. Thesimulation results in the test scenario raised the following twoconcerns: (1) the introduction of SAVs led to increased VMTand, moreover, (2) SAVs attracted public transportation usersrather than private car owners.Snelder et al. [152] developed a simulation framework toassess both direct and indirect impacts of AVs and SAVs ina mixed traffic environment. To capture demand elasticities,the network fundamental diagram was combined with modechoice models. Furthermore, the spatial impact was also mod-eled as an exogenous input to the framework via a percentageof relocated inhabitants per lane use type. The simulation re-sults showed that a shift to SAVs could be expected. However,the improved accessibility for many residents could result in asignificant increase in vehicle trips (and also in VMT), whichmight impose negative effects on traffic condition. Similarconclusions were drawn from the study on the effects of full automation with the possibility of trip chaining of householdtrips, yet in a scenario where most vehicles are still privatelyowned [136].In summary, findings of multiple studies indicate that al-though the introduction of SAVs in the transportation systemmight improve mobility and safety, it could result in enormouschanges of travel behavior, mode choice, car ownership, andpossibly transportation infrastructure and urban form. A holis-tic assess of the impact of the SAV systems on urban mobilityand related social implications might be challenging at themoment as SAVs are still evolving. However, SAV servicecould possibly have negative impact on traffic congestion andbe strongly competitive with public transit without appropriateincentive mechanisms.V. O UTLOOK AND F UTURE D IRECTIONS
A. Concluding Remarks
In this paper, we summarized current research efforts inSAV systems that have been reported in the literature to date.Although the SAV system have many aspects in common withthe conventional carsharing system, the inherent characteristicsof self-driving and self-rebalancing with SAVs further enhancefree-floating carsharing service and increase the stochasticityof the system internally. Externally, the introduction of AVs inthe transportation network could change fundamentally trafficpatterns in the future. The complexity of traffic and urbandynamics, thus, places considerable uncertainty in terms ofboth short-term and long-term impacts of the system [160].The majority of research efforts has considered a systemeither of full SAV penetration rate or without any traffic, andcompared its performance with the conventional mobility sys-tems (in terms of fleet size requirement, energy implications,vehicle miles traveled, passenger travel times, etc). Amongthese research efforts, agent-based modeling is one of themajor approaches to evaluate network performance of a SAVsystem and assess potential impacts of the system. Severalresearch efforts have focused on developing optimizationmodels to address questions (1) “what is the minimum fleetsize to provide a desired level of service?” (2) “What is theoptimal vehicle assignment strategy to minimum passengertravel time?” (3) “What is the optimal vehicle relocationstrategy to minimize the number of rebalancing trips withoutinducing waiting delay?” In general, the SAV system couldbenefit from the cooperative characteristics of the fleet –the connectivity and automation embedded in the systemopen up the opportunities for a central controller to applyoptimal operation strategies to achieve global optimum againstdifferent network design objectives.Although previous research has aimed at enhancing ourunderstanding of the SAV systems, there are still open is-sues to be addressed. For example, most papers considerthe SAV system with fixed stations whereas free floatingSAV systems have not been thoroughly investigated. Withina SAV system, the optimal fleet sizing problem to maintaina minimum required level of service or to ensure a desiredlevel of service is still under-explored. The considerations ofdifferent vehicle assignment and relocation strategies, or the option of ridesharing further increase the complexity of theproblem. So far most papers have applied heuristics for theimplementation of SAVs to solve these problems and focusedmore on assessing potential benefits of a SAV system.
B. Future Research
There are several directions for future research consideringthe gaps in the work reported in the literature to date. Althoughprevious work has addressed the replacement ratio of SAVsto conventional private vehicles, the majority of the resultsare derived with existing demand patterns in an isolatedsystem. The problem of modeling the SAV system withpresence of other transportation modes, as either a complementor competing mode, needs further investigation. Especially,relevant questions still remain unanswered, such as “howis the network performance of such a system in a realistictransportation network?” “How much improvement of levelof service in a transportation network can be achieved withan integrated SAV system?” To address these challenges, itis necessary to study the operational strategies (e.g., optimalfleet size/vehicle assignment/relocation strategy, etc) whichwould yield the minimum and/or desired level of service of thetransportation network. Furthermore, in an environment wheremassive amount of data could be collected from vehicles andinfrastructure, what we used to model as uncertainty becomean additional input. With the advent of information and com-munication technologies, better utilizing available informationfor optimal operational strategies requires novel solutions toreduce dimensions and to overcome issues associated with datain high-dimensional spaces.With all possible mobility service options enabled by CAVs,one particular question that still remains unanswered is “howdemand pattern or travel behavior will eventually change?”With the shared mobility choices (and enhanced conveniencewith SAVs), there is already an evidence of an increase ofinduced demand (e.g., more night travels, or trips shifted fromtransit demand). However, little research has been conductedon investigating the impact of the emerging SAV systemon the vulnerable population, while a systematic frameworkof providing accessibility to a variety of social groups isstill missing. Meanwhile, the nature of self-driving and self-rebalancing of a SAV system also implies potential changes onland use. For example, the implications of a SAV system onurban parking spaces is still under-explored. Thus, the long-term impact of shared mobility system on urban transportationsystems is still an open question.R
EFERENCES[1] A. A. Malikopoulos, “A duality framework for stochastic optimalcontrol of complex systems,”
IEEE Transactions on Automatic Control ,vol. 61, no. 10, pp. 2756–2765, 2016.[2] A. Malikopoulos, “Convergence properties of a computational learningmodel for unknown markov chains,”
J. Dyn. Sys., Meas., Control. , vol.131, no. 3, pp. 041 011–7, 2009.[3] A. Malikopoulos, P. Papalambros, and D. Assanis, “A real-time compu-tational learning model for sequential decision-making problems underuncertainty,”
J. Dyn. Sys., Meas., Control , vol. 131, no. 4, pp. 041 010–8, 2009. [4] S. Shaheen and N. Chan, “Mobility and the sharing economy: Potentialto facilitate the first-and last-mile public transit connections,” BuiltEnvironment , vol. 42, no. 4, pp. 573–588, 2016.[5] S. Shaheen, A. Cohen, and I. Zohdy, “Shared Mobility: CurrentPractices and Guiding Principles,” Federal Highway Administration,Tech. Rep. Washington D.C., 2016.[6] SAE, “Taxonomy and Definitions for Terms Related to Shared Mobilityand Enabling Technologies,” SAE International, Tech. Rep., 2018.[7] A. Cohen and S. Shaheen,
Planning for shared mobility . AmericanPlanning Association, 2018.[8] C. Standing, S. Standing, and S. Biermann, “The implications of thesharing economy for transport,”
Transport Reviews , vol. 39, no. 2, pp.226–242, 2019.[9] S. Feigon and C. Murphy,
Shared mobility and the transformation ofpublic transit . American Public Transportation Association, 2016, no.Project J-11, Task 21.[10] S. Shaheen and A. Cohen, “Is it time for a public transit renaissance?:Navigating travel behavior, technology, and business model shifts ina brave new world,”
Journal of Public Transportation , vol. 21, no. 1,p. 8, 2018.[11] G. G. D. Santos, “One-Way Carsharing Systems: Real-Time Opti-mization of Staff Movements and Operations,” Ph.D. dissertation,Universidade de Lisboa, 2016.[12] E. Martin and S. Shaheen, “Impacts of car2go on vehicle ownership,modal shift, vehicle miles traveled, and greenhouse gas emissions: Ananalysis of five North American cities,”
Transportation SustainabilityResearch Center, UC Berkeley , 2016.[13] ——, “The impact of carsharing on public transit and non-motorizedtravel: An exploration of North American carsharing survey data,”
Energies , vol. 4, no. 11, pp. 2094–2114, 2011.[14] E. Martin and S. A. Shaheen, “Greenhouse gas emissions impacts ofcarsharing in North America,”
Transactions on Intelligent Transporta-tion Systems , vol. 12, no. 4, pp. 1–114, 2011.[15] J. Firnkorn and M. M¨uller, “Free-floating electric carsharing-fleetsin smart cities: The dawning of a post-private car era in urbanenvironments?”
Environmental Science & Policy , vol. 45, pp. 30–40,2015.[16] T. D. Chen and K. M. Kockelman, “Carsharing’s life-cycle impacts onenergy use and greenhouse gas emissions,”
Transportation ResearchPart D: Transport and Environment , vol. 47, pp. 276–284, 2016.[17] M. Barth and S. Shaheen, “Shared-use vehicle systems: Framework forclassifying carsharing, station cars, and combined approaches,”
Trans-portation Research Record: Journal of the Transportation ResearchBoard , vol. 1791, pp. 105–112, 2002.[18] J. Lazarus, S. Shaheen, S. E. Young, D. Fagnant, T. Voege, W. Baum-gardner, J. Fishelson, and J. S. Lott, “Shared automated mobility andpublic transport,” in
Road Vehicle Automation 4 . Springer, 2018, pp.141–161.[19] F. Duarte and C. Ratti, “The impact of autonomous vehicles on cities:A review,”
Journal of Urban Technology , vol. 25, no. 4, pp. 3–18,2018.[20] P. Hancock, I. Nourbakhsh, and J. Stewart, “On the future of transporta-tion in an era of automated and autonomous vehicles,”
Proceedings ofthe National Academy of Sciences , p. 201805770, 2019.[21] S. Shaheen, A. Cohen, B. Yelchuru, S. Sarkhili, B. A. Hamilton et al. , “Mobility on demand operational concept report,” United States.Department of Transportation. Intelligent Transportation . . . , Tech.Rep., 2017.[22] P. Tientrakool, Y. C. Ho, and N. F. Maxemchuk, “Highway capacitybenefits from using vehicle-to-vehicle communication and sensors forcollision avoidance,” in
Vehicular Technology Conference (VTC Fall),2011 . IEEE, 2011, pp. 1–5.[23] S. Shladover, D. Su, and X.-Y. Lu, “Impacts of cooperative adap-tive cruise control on freeway traffic flow,”
Transportation ResearchRecord: Journal of the Transportation Research Board , vol. 2324, no. 1,pp. 63–70, 2012.[24] D. J. Fagnant and K. Kockelman, “Preparing a nation for autonomousvehicles: Opportunities, barriers and policy recommendations,”
Trans-portation Research Part A: Policy and Practice , vol. 77, pp. 167–181,2015.[25] R. Krueger, T. H. Rashidi, and J. M. Rose, “Preferences for sharedautonomous vehicles,”
Transportation research part C: emerging tech-nologies , vol. 69, pp. 343–355, 2016.[26] T. Litman, “Autonomous Vehicle Implementation Predictions: Implica-tions for Transport Planning,” Victoria Transport Policy Institute, Tech.Rep., 2018. [27] A. A. Malikopoulos and L. Zhao, “A closed-form analytical solutionfor optimal coordination of connected and automated vehicles,” in , 2019 (to appear).[28] J. Rios-Torres and A. A. Malikopoulos, “A Survey on Coordinationof Connected and Automated Vehicles at Intersections and Merging atHighway On-Ramps,”
IEEE Transactions on Intelligent TransportationSystems , vol. 18, no. 5, pp. 1066–1077, 2017.[29] A. A. Malikopoulos, C. G. Cassandras, and Y. Zhang, “A decentralizedenergy-optimal control framework for connected automated vehicles atsignal-free intersections,”
Automatica , vol. 93, pp. 244 – 256, 2018.[30] L. Zhao, A. A. Malikopoulos, and J. Rios-Torres, “Optimal control ofconnected and automated vehicles at roundabouts: An investigation ina mixed-traffic environment,” in , 2018, pp. 73–78.[31] A. Stager, L. Bhan, A. A. Malikopoulos, and L. Zhao, “A scaled smartcity for experimental validation of connected and automated vehicles,”in , 2018,pp. 120–135.[32] A. A. Malikopoulos, S. Hong, B. Park, J. Lee, and S. Ryu, “Optimalcontrol for speed harmonization of automated vehicles,”
IEEE Trans-actions on Intelligent Transportation Systems , 2019 (in press).[33] J. Rios-Torres, A. A. Malikopoulos, and P. Pisu, “Online OptimalControl of Connected Vehicles for Efficient Traffic Flow at MergingRoads,” in , 2015, pp. 2432–2437.[34] A. A. Malikopoulos, P. Papalambros, and D. Assanis, “Optimal enginecalibration for individual driving styles,” in
SAE Congress , 2008.[35] A. A. Malikopoulos, P. Y. Papalambros, and D. N. Assanis, “Onlineidentification and stochastic control for autonomous internal combus-tion engines,”
Journal of Dynamic Systems, Measurement, and Control ,vol. 132, no. 2, pp. 024 504–024 504, 2010.[36] A. A. Malikopoulos,
Real-Time, Self-Learning Identification andStochastic Optimal Control of Advanced Powertrain Systems . Pro-Quest, 2008.[37] S. Shaheen and M. Galczynski, “Autonomous carsharing/taxi path-ways,” in
TRB Automated Vehicles Symposium , 2014.[38] R. Zhang, K. Spieser, E. Frazzoli, and M. Pavone, “Models, algorithms,and evaluation for autonomous mobility-on-demand systems,” in
Amer-ican Control Conference (ACC), 2015 . IEEE, 2015, pp. 2573–2587.[39] S. Rodoulis, “The impact of autonomous vehicles on cities,”
Journeys- Sharing Urban Transport Solutions , no. 12, pp. 12–20, 2014.[40] W. Gruel and J. M. Stanford, “Assessing the long-term effects of au-tonomous vehicles: A speculative approach,”
Transportation ResearchProcedia , vol. 13, pp. 18–29, 2016.[41] M. Ohnemus and A. Perl, “Shared autonomous vehicles: Catalyst ofnew mobility for the last mile?”
Built Environment , vol. 42, no. 4, pp.589–602, 2016.[42] A. Stocker and S. Shaheen, “Shared automated vehicle (sav) pilots andautomated vehicle policy in the us: Current and future developments,”in
Road Vehicle Automation 5 . Springer, 2019, pp. 131–147.[43] H. J. Ford, “Shared autonomous taxis: Implementing an efficientalternative to automobile dependency,” B.S. Thesis, Department ofOperations Research and Financial Engineering, Princeton University,2012.[44] L. D. Burns, W. C. Jordan, J. Analytics, and B. A. Scarborough, “Trans-forming Personal Mobility,” The Earth Institute, Columbia University,Tech. Rep., 2012.[45] Z. Chong, B. Qin, T. Bandyopadhyay, T. Wongpiromsarn, B. Reb-samen, P. Dai, E. Rankin, and M. H. Ang, “Autonomy for mobility ondemand,” in
Intelligent Autonomous Systems 12 . Springer, 2013, pp.671–682.[46] D. J. Fagnant and K. M. Kockelman, “The travel and environmentalimplications of shared autonomous vehicles, using agent-based modelscenarios,”
Transportation Research Part C: Emerging Technologies ,vol. 40, pp. 1–13, 2014.[47] J. Bischoff and M. Maciejewski, “Simulation of city-wide replacementof private cars with autonomous taxis in Berlin,”
Procedia ComputerScience , vol. 83, pp. 237–244, 2016.[48] H. Dia and F. Javanshour, “Autonomous shared mobility-on-demand:Melbourne pilot simulation study,”
Transportation Research Procedia ,vol. 22, pp. 285–296, 2017.[49] L. A. Merlin, “Comparing automated shared taxis and conventionalbus transit for a small city,”
Journal of Public Transportation , vol. 20,no. 2, pp. 19–39, 2017.[50] P. M. B¨osch, F. Becker, H. Becker, and K. W. Axhausen, “Cost-basedanalysis of autonomous mobility services,”
Transport Policy , vol. 64,pp. 76–91, 2017. [51] A. Moorthy, R. De Kleine, G. Keoleian, J. Good, and G. Lewis, “Sharedautonomous vehicles as a sustainable solution to the last mile problem:A case study of Ann Arbor-Detroit area,” SAE International Journalof Passenger Cars - Electronic and Electrical Systems , vol. 10, no. 2,pp. 2017–01–1276, 2017.[52] D. Metz, “Developing policy for urban autonomous vehicles: Impacton congestion,”
Urban Science , vol. 2, no. 2, p. 33, 2018.[53] N. Menon, N. Barbour, Y. Zhang, A. R. Pinjari, and F. Mannering,“Shared autonomous vehicles and their potential impacts on householdvehicle ownership: An exploratory empirical assessment,”
InternationalJournal of Sustainable Transportation , vol. 8318, pp. 1–12, 2018.[54] D. F¨oldes and C. Csisz´ar, “Framework for planning the mobility servicebased on autonomous vehicles,” in . IEEE, 2018, pp. 1–6.[55] L. T. Truong, C. De Gruyter, G. Currie, and A. Delbosc, “Estimatingthe trip generation impacts of autonomous vehicles on car travel inVictoria, Australia,”
Transportation , vol. 44, no. 6, pp. 1279–1292,2017.[56] I. P. Tussyadiah, F. J. Zach, and J. Wang, “Attitudes toward autonomouson demand mobility system: The case of self-driving taxi,” in
Infor-mation and Communication Technologies in Tourism 2017 . Springer,2017, pp. 755–766.[57] F. Dandl and K. Bogenberger, “Comparing future autonomous electrictaxis with an existing free-floating carsharing system,”
IEEE Transac-tions on Intelligent Transportation Systems , pp. 1–11, 2018.[58] A. Stocker and S. Shaheen, “Shared automated mobility: early explo-ration and potential impacts,” in
Road Vehicle Automation 4 . Springer,2018, pp. 125–139.[59] ——, “Shared automated vehicles: Review of business models,” in
Roundtable on Cooperative Mobility Systems and Automated Driving ,International Transport Forum Discussion Paper. International Trans-port Forum, 2017.[60] C. Ross and S. Guhathakurta, “Autonomous vehicles and energyimpacts: a scenario analysis,”
Energy Procedia , vol. 143, pp. 47–52,2017.[61] T. D. Chen and K. M. Kockelman, “Management of a shared au-tonomous electric vehicle fleet,”
Transportation Research Record:Journal of the Transportation Research Board , vol. 2572, no. 2572,pp. 37–46, 2016.[62] P. Davidson, A. Spinoulas et al. , “Driving alone versus riding together-how shared autonomous vehicles can change the way we drive,”
Road& Transport Research: A Journal of Australian and New ZealandResearch and Practice , vol. 25, no. 3, p. 51, 2016.[63] D. Jorge and G. Correia, “Carsharing systems demand estimation anddefined operations: A literature review,”
European Journal of Transportand Infrastructure Research , vol. 13, no. 3, pp. 201–220, 2013.[64] N. Agatz, A. Erera, M. Savelsbergh, and X. Wang, “Optimization fordynamic ride-sharing: A review,”
European Journal of OperationalResearch , vol. 223, no. 2, pp. 295–303, 2012.[65] M. Furuhata, M. Dessouky, F. Ord´o˜nez, M.-E. Brunet, X. Wang, andS. Koenig, “Ridesharing: The state-of-the-art and future directions,”
Transportation Research Part B: Methodological , vol. 57, pp. 28–46,2013.[66] G. Brandst¨atter, C. Gambella, M. Leitner, E. Malaguti, F. Masini,J. Puchinger, M. Ruthmair, and D. Vigo, “Overview of optimizationproblems in electric car-sharing system design and management,” in
Dynamic Perspectives on Managerial Decision Making , H. Dawid, Ed.Springer, Cham, 2016, pp. 441–471.[67] P. S. Lavieri, V. M. Garikapati, C. R. Bhat, R. M. Pendyala, S. Astroza,and F. F. Dias, “Modeling individual preferences for ownership andsharing of autonomous vehicle technologies,”
Transportation ResearchRecord: Journal of the Transportation Research Board , no. 2665, pp.1–10, 2017.[68] P. Jittrapirom, V. Caiati, A.-M. Feneri, S. Ebrahimigharehbaghi, M. J.Alonso Gonz´alez, and J. Narayan, “Mobility as a service: A criticalreview of definitions, assessments of schemes, and key challenges,”
Urban Planning , vol. 2, no. 2, pp. 13–25, 2017.[69] R. Utriainen and M. P¨oll¨anen, “Review on mobility as a servicein scientific publications,”
Research in Transportation Business &Management , 2018.[70] M. Hao and T. Yamamoto, “Shared autonomous vehicles : A reviewconsidering car sharing and autonomous vehicles,”
Asian TransportStudies , vol. 5, no. 1, pp. 47–63, 2018.[71] R. Zhang and M. Pavone, “A queueing network approach to theanalysis and control of mobility-on-demand systems,” in
AmericanControl Conference (ACC), 2015 . IEEE, 2015, pp. 4702–4709. [72] ——, “Control of robotic mobility-on-demand systems: A queueing-theoretical perspective,”
The International Journal of Robotics Re-search , vol. 35, no. 3, pp. 186–203, 2016.[73] Spieser, Kevin, K. Ballantyne, Treleaven, R. Zhang, D. Morton, andM. Pavone, “Toward a systematic approach to the design and eval-uation of automated mobility-on-demand systems : A case study inSingapore,” in
Road Vehicle Automation, Lecture Notes in Mobility ,G. Meyer and S. Beike, Eds. Springer International Publishing, 2014,ch. 20, pp. 229–245.[74] M. Pavone, “Autonomous mobility-on-demand systems for future urbanmobility,” in
Autonomes Fahren . Springer, 2015, pp. 399–416.[75] P. M. Boesch and F. Ciari, “Agent-based simulation of autonomouscars,” in
American Control Conference (ACC), 2015 , 2015, pp. 2588–2592.[76] T. D. Chen, K. M. Kockelman, and J. P. Hanna, “Operations of a shared,autonomous, electric vehicle fleet: Implications of vehicle & charginginfrastructure decisions,”
Transportation Research Part A: Policy andPractice , vol. 94, pp. 243–254, 2016.[77] W. Zhang and S. Guhathakurta, “Parking spaces in the age of sharedautonomous vehicles: How much parking will we need and where?”
Transportation Research Record: Journal of the Transportation Re-search Board , no. 2651, pp. 80–91, 2017.[78] W. Zhang, S. Guhathakurta, J. Fang, G. Zhang, W. Zhang, b. S.Guhathakurta, b. G. Zhang, S. Guhathakurta, G. Zhang, and J. Fang,“Exploring the impact of shared autonomous vehicles on urban parkingdemand: An agent-based simulation approach,”
Sustainable Cities andSociety , vol. 19, pp. 34–45, 2015.[79] A. T. Moreno, A. Michalski, C. Llorca, and R. Moeckel, “Sharedautonomous vehicles effect on vehicle-km traveled and average tripduration,”
Journal of Advanced Transportation , vol. 0, no. 0, 2018.[80] C. Brownell and A. Kornhauser, “A driverless alternative: Fleet size andcost requirements for a statewide Autonomous Taxi Network in NewJersey,”
Transportation Research Record: Journal of the TransportationResearch Board , vol. 2416, pp. 73–81, 2014.[81] F. Rossi, R. Zhang, Y. Hindy, and M. Pavone, “Routing autonomousvehicles in congested transportation networks: Structural properties andcoordination algorithms,”
Autonomous Robots , vol. 42, no. 7, pp. 1427–1442, 2018.[82] R. Iglesias, F. Rossi, R. Zhang, and M. Pavone, “A bcmp networkapproach to modeling and controlling autonomous mobility-on-demandsystems,”
The International Journal of Robotics Research , pp. 1–18,2018.[83] B. Jager, F. M. M. Agua, and M. Lienkamp, “Agent-based simulationof a shared, autonomous and electric on-demand mobility solution,”in
Intelligent Transportation Systems (ITSC), 2017 IEEE 20th Interna-tional Conference on . IEEE, 2018, pp. 250–255.[84] A. A. Malikopoulos, “Centralized stochastic optimal control of com-plex systems,” in
Proceedings of the 2015 European Control Confer-ence , 2015, pp. 721–726.[85] A. A. Malikopoulos, V. Maroulas, and J. XIong, “A multiobjectiveoptimization framework for stochastic control of complex systems,”in
Proceedings of the 2015 American Control Conference , 2015, pp.4263–4268.[86] K. A. Marczuk, H. S. S. Hong, C. M. L. Azevedo, M. Adnan, S. D.Pendleton, E. Frazzoli, and D. H. Lee, “Autonomous mobility ondemand in SimMobility: Case study of the central business districtin Singapore,” in
Cybernetics and Intelligent Systems (CIS) and IEEEConference on Robotics, Automation and Mechatronics (RAM), 2015IEEE 7th International Conference on , 2015, pp. 167–172.[87] C. L. Azevedo, K. Marczuk, S. Raveau, H. Soh, M. Adnan, K. Basak,H. Loganathan, N. Deshmunkh, D.-H. Lee, E. Frazzoli, and M. Ben-Akiva, “Microsimulation of demand and supply of autonomous mo-bility on demand,”
Transportation Research Record: Journal of theTransportation Research Board , vol. 2564, pp. 21–30, 2016.[88] D. Kondor, H. Zhang, R. Tachet, P. Santi, and C. Ratti, “Estimatingsavings in parking demand using shared vehicles for home-workcommuting,”
IEEE Transactions on Intelligent Transportation Systems ,2018.[89] K. Winter, O. Cats, G. H. d. A. Correia, and B. van Arem, “Designingan Automated Demand-Responsive Transport System,”
TransportationResearch Record: Journal of the Transportation Research Board , vol.2542, pp. 75–83, 2016.[90] P. M. Boesch, F. Ciari, and K. W. Axhausen, “Autonomous vehiclefleet sizes required to serve different levels of demand,”
TransportationResearch Record: Journal of the Transportation Research Board , vol.2542, pp. 111–119, 2016. [91] D. J. Fagnant, K. M. Kockelman, and P. Bansal, “Operations of sharedautonomous vehicle fleet for Austin, Texas, market,” TransportationResearch Record: Journal of the Transportation Research Board , vol.2536, pp. 98–106, 2015.[92] M. M. Vazifeh, P. Santi, G. Resta, S. H. Strogatz, and C. Ratti,“Addressing the minimum fleet problem in on-demand urban mobility,”
Nature , vol. 557, no. 7706, pp. 534–538, 2018.[93] N. Masoud and R. Jayakrishnan, “Autonomous or driver-less vehicles:Implementation strategies and operational concerns,”
TransportationResearch Part E: Logistics and Transportation Review , vol. 108, pp.179–194, 2017.[94] M. Z. Spivey and W. B. Powell, “The dynamic assignment problem,”
Transportation Science , vol. 38, no. 4, pp. 399–419, 2004.[95] T. ¨Oncan, “A survey of the generalized assignment problem and itsapplications,”
INFOR: Information Systems and Operational Research ,vol. 45, no. 3, pp. 123–141, 2007.[96] V. Pillac, M. Gendreau, C. Gu´eret, and A. L. Medaglia, “A review ofdynamic vehicle routing problems,”
European Journal of OperationalResearch , vol. 225, no. 1, pp. 1–11, 2013.[97] G. La porte, “Fifty years of vehicle routing,”
Transportation Science ,vol. 43, no. 4, pp. 408–416, 2009.[98] J. Ma, X. Li, F. Zhou, and W. Hao, “Designing optimal autonomousvehicle sharing and reservation systems: A linear programming ap-proach,”
Transportation Research Part C: Emerging Technologies ,vol. 84, pp. 124–141, 2017.[99] W. Shen and C. Lopes, “Managing autonomous mobility on demandsystems for better passenger experience,” in
Lecture Notes in ComputerScience (including subseries Lecture Notes in Artificial Intelligence andLecture Notes in Bioinformatics) , 2015, vol. 9387, pp. 20–35.[100] M. Hyland and H. S. Mahmassani, “Dynamic autonomous vehicle fleetoperations: Optimization-based strategies to assign AVs to immediatetraveler demand requests,”
Transportation Research Part C: EmergingTechnologies , vol. 92, pp. 278–297, 2018.[101] J. P. Hanna, M. Albert, D. Chen, and P. Stone, “Minimum cost matchingfor autonomous carsharing,”
IFAC PapersOnLine , vol. 49, no. 15, pp.254–259, 2016.[102] D. Jorge, G. H. Correia, and C. Barnhart, “Comparing optimal re-location operations with simulated relocation policies in one-waycarsharing systems,”
IEEE Transactions on Intelligent TransportationSystems , vol. 15, no. 4, pp. 1667–1675, 2014.[103] S. Weikl and K. Bogenberger, “Relocation strategies and algorithmsfor free-floating car sharing systems,”
IEEE Intelligent TransportationSystems Magazine , vol. 5, no. 4, pp. 100–111, 2013.[104] A. B. Brendel, S. Lichtenberg, I. Nastjuk, and L. M. Kolbe, “Adaptingcarsharing vehicle relocation strategies for shared autonomous electricvehicle services,” in , Seoul, South Korea, 2017, pp. 1–20.[105] G. G. D. Santos and G. H. de Almeida Correia, “Finding the rele-vance of staff-based vehicle relocations in one-way carsharing systemsthrough the use of a simulation-based optimization tool,”
Journal ofIntelligent Transportation Systems , pp. 1–22, 2019.[106] R. Zhang, F. Rossi, and M. Pavone, “Model predictive control ofautonomous mobility-on-demand systems,” in . IEEE, 2016, pp.1382–1389.[107] J. Wen, J. Zhao, and P. Jaillet, “Rebalancing shared mobility-on-demand systems: A reinforcement learning approach,” in
IntelligentTransportation Systems (ITSC), 2017 IEEE 20th International Confer-ence on . IEEE, 2017, pp. 220–225.[108] K. Spieser, W. Gruel, and E. Frazzoli, “Shared-vehicle mobility-on-demand systems: a fleet operator’s guide to rebalancing emptyvehicles,” in
Transportation Research Board 95th Annual Meeting ,2015.[109] M. Pavone, S. L. Smith, E. Frazzoli, and D. Rus, “Robotic loadbalancing for mobility-on-demand systems,”
The International Journalof Robotics Research , vol. 31, no. 7, pp. 839–854, 2012.[110] S. H¨orl, C. Ruch, F. Becker, E. Frazzoli, and K. W. Axhausen, “Fleetoperational policies for automated mobility: A simulation assessmentfor zurich,”
Transportation Research Part C: Emerging Technologies ,vol. 102, pp. 20–31, 2019.[111] S. Zhu and A. L. Kornhauser, “The interplay between fleet size,level of service and empty vehicle repositioning strategies in large-scale, shared-ride autonomous taxi mobility-on-demand scenarios,” in
Transportation Research Board 96th Annual Meeting , 2017, pp. 1–14.[112] K. A. Marczuk, H. S. Soh, C. M. Azevedo, D.-H. Lee, and E. Frazzoli,“Simulation framework for rebalancing of autonomous mobility on demand systems,” in
MATEC Web of Conferences , M. Figueira andZ. Guo, Eds., vol. 81. EDP Sciences, 2016.[113] K. Winter, O. Cats, B. Van Arem, and K. Martens, “Impact of relocationstrategies for a fleet of shared automated vehicles on service efficiency,effectiveness and externalities,” in
Models and Technologies for Intel-ligent Transportation Systems (MT-ITS), 2017 5th IEEE InternationalConference on . IEEE, 2017, pp. 844–849.[114] M. Winter, O. Cats, K. Martens, and B. van Arem, “Relocatingstrategies under parking constraints for a fleet of shared automatedvehicles,” hEART 2018 , 2018.[115] N. Kang, F. M. Feinberg, and P. Y. Papalambros, “Autonomous electricvehicle sharing system design,”
Journal of Mechanical Design , vol.139, no. 1, 2016.[116] B. Loeb, K. M. Kockelman, and J. Liu, “Shared autonomous electricvehicle (SAEV) operations across the Austin, Texas network withcharging infrastructure decisions,”
Transportation Research Part C:Emerging Technologies , vol. 89, pp. 222–233, 2018.[117] B. Loeb and K. M. Kockelman, “Fleet performance and cost evaluationof a shared autonomous electric vehicle (saev) fleet: A case study foraustin, texas,”
Transportation Research Part A: Policy and Practice ,vol. 121, pp. 374 – 385, 2019.[118] G. S. Bauer, J. B. Greenblatt, and B. F. Gerke, “Cost, energy, andenvironmental impact of automated electric taxi fleets in manhattan,”
Environmental science & technology , vol. 52, no. 8, pp. 4920–4928,2018.[119] R. Iacobucci, B. McLellan, and T. Tezuka, “Modeling shared au-tonomous electric vehicles: Potential for transport and power gridintegration,”
Energy , vol. 158, pp. 148–163, 2018.[120] ——, “Optimization of shared autonomous electric vehicles operationswith charge scheduling and vehicle-to-grid,”
Transportation ResearchPart C: Emerging Technologies , vol. 100, pp. 34–52, 2019.[121] H. Miao, H. Jia, J. Li, and T. Z. Qiu, “Autonomous connected electricvehicle (acev)-based car-sharing system modeling and optimal plan-ning: A unified two-stage multi-objective optimization methodology,”
Energy , vol. 169, pp. 797–818, 2019.[122] V. Oldenbroek, L. A. Verhoef, and A. J. van Wijk, “Fuel cell electricvehicle as a power plant: Fully renewable integrated transport andenergy system design and analysis for smart city areas,”
InternationalJournal of Hydrogen Energy , vol. 42, no. 12, pp. 8166 – 8196, 2017.[123] S. Samaranayake, K. Spieser, H. Guntha, and E. Frazzoli, “Ridepoolingwith trip-chaining in a shared-vehicle mobility-on-demand system,” in , 2017, pp. 1–7.[124] M. W. Levin, K. M. Kockelman, S. D. Boyles, and T. Li, “A generalframework for modeling shared autonomous vehicles with dynamicnetwork-loading and dynamic ride-sharing application,”
Computers,Environment and Urban Systems , vol. 64, pp. 373–383, 2017.[125] M. F. Hyland and H. S. Mahmassani, “Sharing is caring: Dynamicautonomous vehicle fleet operations under demand surges,” in
Trans-portation Research Board 97th Annual Meeting , 2018.[126] D. J. Fagnant and K. M. Kockelman, “Dynamic ride-sharing and fleetsizing for a system of shared autonomous vehicles in Austin, Texas,”
Transportation , vol. 45, no. 1, pp. 143–158, 2018.[127] M. Tsao, D. Milojevic, C. Ruch, M. Salazar Villalon, E. Frazzoli,and M. Pavone, “Model predictive control of ride-sharing autonomousmobility-on-demand systems,” in , 2016, pp. 1382–1389.[128] W. Burghout, P.-J. Rigole, and I. Andreasson, “Impacts of sharedautonomous taxis in a metropolitan area,” in
Transportation ResearchBoard 94th Annual Meeting , 2015.[129] K. M. Gurumurthy and K. M. Kockelman, “Analyzing the dynamicride-sharing potential for shared autonomous vehicle fleets using cell-phone data from orlando, florida,”
Computers, Environment and UrbanSystems , vol. 71, pp. 177 – 185, 2018.[130] R. Iglesias, F. Rossi, K. Wang, D. Hallac, J. Leskovec, and M. Pavone,“Data-driven model predictive control of autonomous mobility-on-demand systems,” in . IEEE, 2018, pp. 1–7.[131] M. W. Levin, “Congestion-aware system optimal route choice forshared autonomous vehicles,”
Transportation Research Part C: Emerg-ing Technologies , vol. 82, pp. 229–247, 2017.[132] X. Liang, G. H. De Almeida Correia, and B. van Arem, “An optimiza-tion model for vehicle routing of automated taxi trips with dynamictravel times,”
Transportation Research Procedia , vol. 27, pp. 736–743,2017.[133] X. Liang, G. H. de Almeida Correia, and B. van Arem, “Applying amodel for trip assignment and dynamic routing of automated taxis with congestion: system performance in the City of Delft, The Netherlands,” Transportation Research Record , 2018.[134] X. Liang, G. H. d. A. Correia, and B. van Arem, “Optimizing theservice area and trip selection of an electric automated taxi systemused for the last mile of train trips,”
Transportation Research Part E:Logistics and Transportation Review , vol. 93, pp. 115–129, 2016.[135] X. Liang, K. An, G. Homem de Almeida Correia, and B. van Arem,“An optimization model of automated taxis in trip assignment underelastic demand for the first/last mile problem,” hEART 2018 , 2018.[136] G. H. de Almeida Correia and B. van Arem, “Solving the user optimumprivately owned automated vehicles assignment problem (uo-poavap):A model to explore the impacts of self-driving vehicles on urbanmobility,”
Transportation Research Part B: Methodological , vol. 87,pp. 64–88, 2016.[137] P.-J. Rigole, “Study of a shared autonomous vehicles based mobilitysolution in Stockholm,” M.S. Thesis, Division of Industrial Ecology,Royal Institute of Technology, 2014.[138] H. K. R. d. F. Pinto, M. F. Hyland, ˙I. ¨O. Verbas, H. S. Mahmassani, andT. R. Board, “Integrated mode choice and dynamic traveler assignment-simulation framework to assess the impact of a suburban first-mileshared autonomous vehicle fleet service on transit demand,”
Trans-portation Research Record: Journal of the Transportation ResearchBoard , 2018.[139] D. Fiedler, M. Cap, and M. Certicky, “Impact of mobility-on-demandon traffic congestion: Simulation-based study,” in
Intelligent Trans-portation Systems (ITSC), 2017 IEEE 20th International Conferenceon . IEEE, 2018, pp. 1–6.[140] A. Scheltes and G. H. de Almeida Correia, “Exploring the use ofautomated vehicles as last mile connection of train trips through anagent-based simulation model: An application to Delft, Netherlands,”
International Journal of Transportation Science and Technology , vol. 6,no. 1, pp. 28–41, 2017.[141] J. Farhan and T. D. Chen, “Impact of ridesharing on operationalefficiency of shared autonomous electric vehicle fleet,”
TransportationResearch Part C: Emerging Technologies , vol. 93, pp. 310–321, 2018.[142] W. Zhang, S. Guhathakurta, J. Fang, and G. Zhang, “The performanceand benefits of a shared autonomous vehicles based dynamic rideshar-ing system : An agent-based simulation approach,” in
TransportationResearch Board 94th Annual Meeting , 2015.[143] J. Zachariah, J. Gao, A. Kornhauser, and T. Mufti, “Uncongestedmobility for all: A proposal for an area wide autonomous taxi system inNew Jersey,” in
Transportation Research Board 93rd Annual Meeting ,2014, pp. 1–14.[144] M. Lu, M. Taiebat, M. Xu, and S.-C. Hsu, “Multiagent spatial simu-lation of autonomous taxis for urban commute: Travel economics andenvironmental impacts,”
Journal of Urban Planning and Development ,vol. 144, no. 4, 2018.[145] W. Shen, C. V. Lopes, and J. W. Crandall, “An online mechanism forridesharing in autonomous mobility-on-demand systems,” in
Twenty-Fifth International Joint Conference on Artificial Intelligence (IJCAI-16) , 2016, pp. 475–481.[146] Y. Shen, H. Zhang, and J. Zhao, “Integrating shared autonomousvehicle in public transportation system: A supply-side simulation ofthe first-mile service in singapore,”
Transportation Research Part A:Policy and Practice , vol. 113, pp. 125–136, 2018.[147] R. Chen and C. G. Cassandras, “Optimization of ride sharing sys-tems using event-driven receding horizon control,” arXiv preprintarXiv:1901.01919 , 2019.[148] S. H¨orl, “Implementation of an autonomous taxi service in a multi-modal traffic simulation using MATSim,” Ph.D. dissertation, ChalmersUniversity of Technology, 2016.[149] J. Liu, K. M. Kockelman, P. M. Boesch, and F. Ciari, “Tracking asystem of shared autonomous vehicles across the Austin, Texas networkusing agent-based simulation,”
Transportation , vol. 44, no. 6, pp. 1261–1278, 2017.[150] L. M. Martinez and J. M. Viegas, “Assessing the impacts of deployinga shared self-driving urban mobility system: An agent-based modelapplied to the city of Lisbon, Portugal,”
International Journal ofTransportation Science and Technology , vol. 6, no. 1, pp. 13–27, 2017.[151] Y. Shen, H. Zhang, and J. Zhao, “Embedding autonomous vehiclesharing in a public transit system: Example of last-mile problem,” in
Transportation Research Board 96th Annual Meeting , 2017.[152] M. Snelder, I. Wilmink, J. van der Gun, H. J. Bergveld, P. Hoseini,and B. van Arem, “Mobility impacts of automated driving and sharedmobility–explorative model and case study of the province of north-holland,” in
Transportation Research Board 98th Annual Meeting ,2019. [153] J. Wen, Y. X. Chen, N. Nassir, and J. Zhao, “Transit-oriented au-tonomous vehicle operation with integrated demand-supply interac-tion,”
Transportation Research Part C: Emerging Technologies , vol. 97,pp. 216–234, 2018.[154] R. Basu, A. Araldo, A. P. Akkinepally, B. H. N. Biran, K. Basak,R. Seshadri, N. Deshmukh, N. Kumar, C. L. Azevedo, and M. Ben-Akiva, “Automated mobility-on-demand vs. mass transit: A multi-modal activity-driven agent-based simulation approach,”
Transporta-tion Research Record , vol. 0, no. 0, 2018.[155] R. Lamotte, A. De Palma, and N. Geroliminis, “On the use ofreservation-based autonomous vehicles for demand management,”
Transportation Research Part B: Methodological , vol. 99, pp. 205–227,2017.[156] M. Maciejewski and J. Bischoff, “Congestion effects of autonomoustaxi fleets,”
Transport , pp. 1–10, 2017.[157] B. Remer and A. A. Malikopoulos, “The multi-objective dynamictraveling salesman problem: Last mile delivery with unmanned aerialvehicles assistance,” in
American Control Conference , 2019 (to appear).[158] R. Boersma, B. van Arem, and F. Rieck, “Application of driverlesselectric automated shuttles for public transport in villages: The case ofappelscha,”
World Electric vehicle journal , vol. 9, no. 1, p. 15, 2018.[159] Z. Wadud, D. MacKenzie, and P. Leiby, “Help or hindrance? thetravel, energy and carbon impacts of highly automated vehicles,”
Transportation Research Part A: Policy and Practice , vol. 86, pp. 1–18,2016.[160] D. Milakis, B. Van Arem, and B. Van Wee, “Policy and society relatedimplications of automated driving: A review of literature and directionsfor future research,”
Journal of Intelligent Transportation Systems ,vol. 21, no. 4, pp. 324–348, 2017.
Liuhui Zhao (M2017) received the B.S. degreein Resources Science and Technology from BeijingNormal University, Beijing, China, in 2009, theM.S. degree from Department of Geography at theUniversity of Alabama in 2011, and the Ph.D. de-gree in Transportation Engineering from New JerseyInstitute of Technology in 2016. She is currentlya Postdoctoral Researcher in the Information andDecision Science (IDS) Laboratory at the Universityof Delaware leading research projects on emergingtransportation systems. She has participated in vari-ous research projects on connected automated vehicles, intelligent transporta-tion systems, traffic and transit operations. Her research interests lie withinthe areas of intelligent transportation systems, shared mobility, and connectedautomated vehicles.