Self-driving cars enable dynamic shared mobility, where customers are independent of schedules and fixed stops. This study aims to investigate the potential effects shared mobility can have on future transportation. We simulate multiple scenarios to analyze the effects different service designs might have on vehicle kilometers, on the required number of shared vehicles, on the potential replacement of private cars, and on service metrics such as waiting times, travel times, and detour levels. To demonstrate how simulation can be used to analyze future mobility, we present a case study of the city of Gothenburg in Sweden, where we model travel demand in the morning hours of a workday. The results show that a significant decrease of vehicle kilometers can be achieved if all private car trips are replaced by rideshare and that shared vehicles can potentially replace at least 5 private cars during the morning peak.
Hinweise
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
1 Introduction
The shift to self-driving cars and autonomous mobility is ongoing and will presumably have a great influence on future transportation. A reasonable assumption is that the removed costs for drivers will boost new services, such as on demand services. This will most likely not only affect private road users but also public transportation. Already today, there is an increasing demand in dynamic transport, which is independent of schedules or fixed stops, i.e., Demand Responsive Transport [1, 2]. This transport mode allows customers to request trips according to their individual needs, which are flexible in desired departure time as well as Pick-Up and Drop-Off location (PUDO). Based on these requests, the service provider plans individual journeys for the travelers. To further increase the efficiency of such a system, the possibility of ridesharing may be used, where passengers with different origins and destinations are going together in the same vehicle for parts of the journey [3, 4]. Even though dynamic self-driving ridesharing services are not yet fully ready for the market, there exist several pilot studies where such services were tested, indicating a great potential for future mobility [5].
It is in general unclear how and to what extent people will use shared self-driving vehicles in the future, since it enables new transport opportunities whose effects are difficult to anticipate. Additionally, incentives and regulation may shift or force travel behavior to a higher degree of usage of ride-shared vehicles. Due to the large number of influencing factors, it is challenging to investigate how autonomous shared mobility might affect future transportation. One approach to analyze the potential effects of different services and regulations is the use of computer simulation. Building a simulation model of a traffic system allows for conducting what-if analyses of possible scenarios, which can provide decision-makers with new insights on how different settings might, for instance, affect traffic volume or travel times. To this end, a variety of simulation studies has been conducted in different cities with the aim of investigating the effects of modern shared mobility systems, e.g., in Oslo [6], Lisbon [7, 8], and Helsinki [9]. These existing studies provide valuable insights on how the general shift to autonomous vehicles might affect mobility in cities.
Anzeige
The goal of this study is to complement earlier studies and to better understand the consequences and influencing factors of shared mobility. We investigate the shift to shared mobility under different circumstances, i.e., for different traveler groups, parts of the city, and road types as well as factors that might improve ridesharing (e.g., allocation of PUDOs and holding areas). This study does not aim to predict the changes in future travel behavior or mode choices as such. Instead, our intention is to conduct a what-if analysis to investigate the potential effects of a future shift to shared mobility, assuming that different push and pull measures are successfully implemented to reduce travel by private car and to govern the transition to shared mobility (e.g., taxes, fees, mobility management campaigns, or reducing number of parking spaces) [10]. The study aims to capture the possible consequences of a potentially extensive usage of shared self-driving vehicles and to better understand how different factors affect mobility, e.g., varying levels of service of the shared vehicles and different types of mobility users (i.e., current public transport or private car users). We simulate different scenarios and analyze the effects different service designs might have on vehicle kilometers (vkm), on the required fleet size, on the potential replacement of private cars (non-shared), and on service metrics such as waiting times, travel times, and detour levels. We also simulate the use of holding areas where shared vehicles wait when idling between trips.
To demonstrate how simulation can be used to analyze future mobility, we present a case study of the city of Gothenburg in Sweden, where we model the (origin–destination-based) travel demand in the morning peak hours of a workday. This study complements and extends on earlier simulation studies on shared mobility systems by providing more detailed results on the effects different scenarios have on the overall mobility and traffic but also on specific road types (e.g., central roads or residential areas). We investigate different scenarios of private car and public transport users shifting to carshare and rideshare but also the effects of other parameters such as distance between PUDOs and detour tolerance. Moreover, the presented study includes a more fine-grained representation of origins and destinations and a more realistic and efficient fleet management. To this end, we also provide an overview of related studies and compare our results to those of the other studies.
We use our simulation model to investigate the potential effects the shift from private cars and traditional public transport to carsharing and ridesharing can have on society and future transportation systems. Specific questions that we address are, for instance:
How does the traffic volume change and which road types are mostly affected?
How do carsharing and ridesharing affect waiting, detour, and travel times of users?
Where should PUDOs be located and what distance between the PUDOs is suitable?
How many holding areas are required and where should they be located?
How many shared vehicles are required to satisfy the transport demand?
The article is structured as follows: Section 2 provides a background shared mobility on demand and presents related simulation studies. In Section 3, we present our simulation model as well as the mobility simulator we used for our study. Section 4 introduces the case study of the Gothenburg – Mölndal – Partille area in terms of the data we used to generate the demand for shared mobility as well as the scenarios we simulated. In Section 5, we present the results of the simulations and provide a discussion of these results in Section 6, where we also compare them to the results of related studies. Finally, in Section 7, we draw conclusions and discuss future work.
Anzeige
2 Background
Mobility as a Service (MaaS) is gaining popularity, where transport services are customized to the individual travelers’ demand [11]. In the long term, it is possible that this new form of transportation will replace ownership-based transportation systems, possibly even resulting in the substitution of private cars. Different concepts exist for providing MaaS, whose feasibility has been evaluated in numerous studies. This chapter provides an overview of concepts and studies with focus on on-demand services.
2.1 Shared Mobility on Demand
To counteract the increasing traffic volume and resulting emissions, there is a growing interest in more sustainable transport solutions [12]. The goal of these solutions is often the reduction of private car use, which are a significant part of the current transportation system. Private cars are on average parked 95% of their lifetime [13] and that millennials, i.e., those that born in the 1980s and 90 s, are less interested in owning a car [14].
There are two common approaches for shared mobility, carsharing and ridesharing, where travelers have access to multiple shared vehicles, which they can use for their trips [15]. In contrast to private cars, travelers do not own the cars and avoid ownership responsibilities, e.g., service, insurance, or parking. Carsharing is similar to traditional taxis, where only one passenger or group of passengers traveling together is served at a time. Ridesharing, often also referred to as ridepooling, tries to further increase the utilization of vehicles and enables different travelers to combine their trips. Yet, this might result in increased travel time due to deviations from the direct route for picking up or dropping off other travelers.
According to Furuhata et al. [4, p. 28], ridesharing combines “flexibility and speed of private cars with the reduced cost of fixed-line systems”. Both travel modes can provide flexible mobility without owning a vehicle, yet, they might require travelers to accept inconveniences such as longer walking distances to the trip’s starting point or detours due to other travelers [4].
2.2 Simulation Studies of Shared Mobility
For policy and decision makers, the trend from private towards shared mobility implies challenges regarding the planning and development of cities. It includes the adaption of the traffic system and road network but also issuing laws and policies that govern the transition. Numerous simulation studies have been conducted in different cities that investigate the effects shared mobility can have on the transport system. Based on a model that explicitly represents the inhabitants of the city, their travel demand, and the road network, the effects can be investigated when different groups of travelers or a certain share of travelers decides to use shared mobility.
The International Transport Forum (ITF) has conducted simulation studies on shared mobility in Lisbon (Portugal), Helsinki (Finland), Auckland (New Zealand), Dublin (Ireland), and Lyon (France) (see Table 1). These studies use agent-based models with a synthetic population of travelers, where trip demand is stochastically generated using an travel simulation model [16], which is based on travel survey and census data of the respective region. Travelers request trips from one point to another and the dispatcher optimizes the routing of the vehicle taking different parameters into account. The traveler then walks to the pickup location, boards the vehicle, and exits the system when arriving the destination. The simulation takes different means of transportation into consideration such as walking, biking, private cars, and public transport but also shared modes of transportation. For shared mobility, the travelers request a trip from and to a specific location with a desired departure or arrival time. The centralized dispatcher then tries to plan and optimize the journey under due consideration of other travelers that also requested journeys, which can be combined in terms of ridesharing.
Table 1
Overview of related simulation studies on shared mobility with the Gothenburg area studied here as reference
Year
Area (km2)
Population
Daily trips (base scenario)
Lisbon
2015
85
565 000
1 200 000
Helsinki
2017
770
1 089 000
3 100 000
Auckland
2017
2 200
1 300 000
4 500 000
Dublin
2018
7 000
1 800 000
4 600 000
Oslo
2019
5 400
1 300 000
401 000*
Lyon
2020
534
1 340 000
3 200 000
Gothenburg
2021
200
700 000
1 600 000
*trips between 6 and 10 a.m
The Lisbon study [7] was the first study of this series but also the smallest of the studies with respect to the investigated area and the size of the population of potential travelers. While the other studies analyze the entire metropolitan region of the respective city, the Lisbon study only considers the city itself, resulting in a denser demand of trips, which also facilitates the provision of high-quality ridesharing services. The authors show that the number of vehicles can be reduced by 95% for some scenarios and that the idle time of vehicles can be decreased from 96 to 27% in case all car and public transport users shift to ridesharing. Still, vkm increase by up to 200% depending on the investigated scenario.
A similar study was conducted in Helsinki [9]. While the model was mostly similar, some extensions were made, e.g., the capacity of vehicles and different booking options. Carsharing allows for the real-time booking of door-to-door trips whereas ridesharing requires the traveler to book at least 30 min in advance with a walking distance of up to 400 m to the nearest PUDO. The potential for reducing the number of vehicles corresponds to the findings from the Lisbon study, yet, the reduction in vkm compared to the baseline with private cars is considerably lower (33%). The authors argue that the shift to shared mobility might increase the 45-min accessibility to employment from 40% to up to 56% due to decreased travel times for former PT users.
Both the Auckland study [17] and the Dublin study [18] use the same vehicle configurations, booking options, and service provision as in Helsinki. Yet, these studies also investigate the effects of Low Emission Zones (LEZ) where restrictions of car usage apply as well as the removal of less-frequented bus lines. The studies show that LEZ can indeed reduce congestion, yet, result in a less evenly distributed service provision. Moreover, the studies underline the suitability of shared mobility as feeder to existing PT lines as well as the inefficiency of vehicles with higher capacity in areas with lower demand density. The studies indicate an up to 50% reduction in vkm when replacing all private car trips with shared mobility.
The latest study from this series concerns Lyon [19], where carpooling with private vehicles is investigated in addition to carsharing and ridesharing services. Carpooling allows owners of private cars to meet up with other travelers at designated parking lots to then jointly drive to a location that is no more than 1 km from the desired destination of the traveler. The study shows that most of the travelers (75%) will experience short waiting times and detours when using shared mobility modes and that shifting all trips to shared mobility might lead to a reduction of 94% in the total vehicle fleet.
A simulation study for Oslo was conducted by COWI & PTV [6]. It investigates the effects of both private car and PT users in the Oslo region switching to shared mobility. Like in the other studies, vehicles of different size are used to offer shared mobility on a junction-to-junction basis with real-time booking. The greatest reduction in fleet size was achieved for the scenario where all car travelers use ridesharing (-93%) but even in all other simulated scenarios, the number of vehicles required by shared mobility services is significantly lower than the original volume of traffic. In terms of vkm, however, a decrease was only observed when shifting from private cars to ridesharing (-16%), which can be explained by an increasing number of empty trips. From the travelers’ perspective, a full replacement of private car trips with ridesharing will result in an increased travel time (67%), which also includes resulting waiting times, and a slightly increased travel distance (7.7%).
Besides, a series of smaller studies exist, that cover individual aspect of introducing shared mobility services, e.g., in Berlin [20], Barcelona [21], Los Angeles [22], and Stuttgart [23]. However, some of these studies have limited relevance compared to our study, for instance, as they do not consider public transport, only provide shared mobility to a small amount of the travelers, or as they do not provide outputs that are required for a comparison, e.g., vehicle kilometers or travel time. Hence, we do not include these studies into our comparison of results.
3 Simulating the Impact of Shared Mobility on Demand
The purpose of this study is to investigate the potential effects shared mobility with self-driving vehicles have on future transportation in terms of, e.g., vehicle kilometers, traffic volume, and travel times. Hence, shared mobility concepts such as carsharing and ridesharing need to be modeled at individual level. This includes modelling of each traveler’s demand for mobility as well as serving of this demand by individual vehicles.
The simulation model we present in this chapter consists of a macroscopic representation of the traffic system, i.e., the traffic network and the zone-to-zone trip demand, as well as an individual-based (microscopic) simulation of fleet dispatching to meet each travelers’ on-demand travel requests (see Fig. 1).
Fig. 1
Overview of the simulation model
×
3.1 Modelling the Traffic System
To investigate the effects of shared mobility concepts, the traffic system in question needs to be modeled first. Required inputs include, for instance, the service area, i.e., the traffic network of the investigated service area, and the locations from and to which travelers can request trips. Another essential aspect is the trip demand, i.e., the travel habits of travelers. The acquisition of individual-based data is challenging, and most traffic models use accumulated zone-to-zone trip data, i.e., an origin–destination (OD) matrix.
The OD matrices we use in this model come from Gothenburg’s Strategic Traffic Model (GSM; Göteborgs Strategiska Trafikmodell) [24]. The GSM is a multi-modal travel demand model used by the city of Gothenburg to analyze and plan future traffic development. The model has been developed using different data sources and its plausibility was validated by traffic experts from the municipality as well as using real traffic data. Hence, we assume that the travel demand from the OD-matrices from the GSM is credible, which we use as input for our simulation model.
The GSM uses night-time and day-time population as well as land use data to estimate the production and attraction of each zone. Moreover, household survey data and travel survey data on a zone level is used to gather information, amongst others, on average number of trips per day, number of cars per household, and mode share.
As some travelers combine trips of different purposes, e.g., going to work and doing grocery shopping on the way home, the OD matrices we use as input for our simulations are estimated based on trip chains. These matrices differ between travel modes and are calculated using a Multinomial Logit model with utility functions for different traveler groups. The mode choice is based on generalized utility costs for using different transport modes and weighting parameters derived from a household survey. This includes, e.g., in-vehicle travel time, waiting time, PT fare, number of PT transfers, and parking costs. Common examples of activity chains are home–school–work or work–shop–home.
Based on all these factors and together with the travel supply options of the different transport modes, travel demand and mode choice for each zone are calculated in an iterative process until the generated traffic volumes correspond to actual traffic counts. The resulting trips are stored in an OD-matrix for each mode of transport.
We use these OD matrices as inputs for the shared mobility simulations presented in this article. Moreover, they serve as base scenario in the traffic analysis we present and allow for comparing the simulation results. In our study, these OD matrices will not be changed to consider changes in travel preferences, e.g., due to the introduction of new services, or other traffic related phenomena such as congestion.
Travelers are assumed to be distributed across the zone they travel from. OD matrices, however, only provide aggregated information on the number of occurring trips from and to each zone and do not allocate these trips to specific addresses or locations. The occurring demand, thus, needs to be distributed in a more realistic way to model service provision by autonomous vehicles in more detail. Travelers most likely request trips from and to a specific address, e.g., their home. In the simulation model, such points from and to which travelers can request their trips are referred to as access and egress nodes. For each simulated traveler and in accordance with the OD matrix, a pair of access and egress nodes is generated from where they request their trip and where they want to travel to. Moreover, the exact point in time for which the traveler requests the trip is randomly determined in accordance with the hourly distribution of zone-level travel demand.
A great number of access and egress nodes might exist, which makes it difficult for the service provider to efficiently plan and offer trips due to great number of potential origins and destinations, which also result in increased detours. Thus, it is feasible to define specific Pick-Up and Drop-Off locations (PUDOs), virtual stops along the road network from and to which passengers can request trips. By varying the distance between PUDOs, different service degrees can be simulated, reaching from door-to-door service to a more wide-meshed network of stops. Thus, simulation can be used to investigate how different strategies for placing PUDOs affect both service provision and customer satisfaction and the best suited configuration of PUDOs can be identified [25].
In the Oslo study, criteria for the generation of suitable PUDO locations include roads that allow car traffic, that do not consist of more than two lanes, that do not have a speed limit of more than 60 km/h, and that are not located in tunnels or on overpasses. After identifying all potential locations, nodes can be removed iteratively until a certain minimum distance between the PUDOs is achieved. Once a traveler requests a trip, the pick-up will be planned from the PUDO that is closest to its origin (access node) to the PUDO closest to the destination (egress node).
For our simulations, we use PTV Visum to model the traffic system and the trip demand, a macroscopic traffic analysis tool for both car traffic and public transport. It does not consider individual travelers and road users but instead models traffic on a more aggregated level in terms of flow and density. This allows for the more efficient analysis of greater traffic volume and larger areas, e.g., for investigating the effects of changes in road infrastructure or of new public transport lines. Moreover, PTV Visum allows for defining access and egress nodes, holding areas, and PUDOs.
3.2 Modelling Shared Mobility
To simulate the effects of shared mobility, the zone-to-zone trip demand needs to be transformed into the trips requested by individual travelers. Accordingly, the execution of the simulation consists of different steps. First, based on the OD-matrix, trip demand is generated in the respective zones. For each traveler, the access and egress nodes are randomly chosen within the zones as well as the departure time of the trip. The journey of the traveler starts by leaving the access node and walking to the closest PUDO where he or she waits for the arrival of the vehicle. Starting from this origin PUDO, a trip is planned taking the travelers’ detour acceptance and requests of other travelers into account. The goal of the simulation is to optimize the utilization of ridesharing vehicles while also ensuring a high quality of service for the travelers. After being dropped off at the desired destination PUDO, the traveler walks the remaining distance to the egress node.
There are parameters of the simulation that are specific to shared mobility. This includes, for instance, the configuration of the vehicles, i.e., how many seats they have. To meet the demand for trips, service providers can deploy different configurations of vehicles and with respect to route planning, the capacity of the vehicles is an important factor. On heavily frequented routes, larger vehicles might allow for a more efficient service provision.
Sustainable shared mobility requires increasing the occupancy of the vehicles, which might result in longer individual travel times and distances compared to direct trips [26]. As seen in today’s public transport system, there exists different approaches for increasing the travelers’ willingness to accept detours. This includes different push and pull measures to discourage private mobility and to promote shared mobility, e.g., economic incentives such as discounts compared to private rides, public awareness campaigns, parking restrictions, speed reductions, or road taxes. Viergutz & Krajzewicz compare the resulting travel time ratios of private car and public transport use between different European cities indicating an average ratio of 1.25 to 1.71 in medium sized cities [27]. The study by König & Grippenkoven shows how different detour factors can be incentivized through discounts [26]. Hence, in our model, we assume travelers having a certain detour acceptance, which might be the results of different push and pull measures put in place.
The most suitable capacity of the vehicles also depends on whether they are used for carsharing or ridesharing, where smaller vehicles might be sufficient for most carsharing trips. In the simulation study conducted in Oslo, three types of vehicles were used: shared cars with space for 4, 6, and up to 20 passengers. In contrast to this, the authors of the Helsinki study simulated two types of vehicles, shared vehicles with capacity for 8 or 16 passengers. Ultimately, it is up to the modeler to define different vehicle configurations and to analyze their suitability.
Besides vehicle capacity, there are also other parameters that are relevant for the design of carsharing and ridesharing services, e.g., the possibility of pre-booking a trip, the maximum waiting and detour time that is accepted by the travelers, boarding and unboarding times, and fleet size. These parameters can be used to define different scenarios that can be simulated.
Accordingly, we import the network-based traffic system model from PTV Visum to PTV MaaS Modeller, which allows us to measure the effects of shared mobility by simulating and optimizing fleet dispatching on a microscopic level to meet on-demand travel requests. In PTV MaaS Modeller, shared mobility parameters can be set, e.g., acceptable waiting time, accepted detour factor, prebooking time, and available fleet size. Based on these parameters and on the individual trip demand of travelers from the PUDO that is closest to their access node to the PUDO that is closest to their egress node, PTV MaaS Modeller dispatches and routes vehicles such that all travelers are served with a ride while optimizing the fleet size and minimizing the total vkm. For carsharing, this is achieved by adding suitable travelers to a trip in case the resulting detour does not exceed the travelers’ detour tolerance. Through this, we can efficiently simulate MaaS concepts on a large scale, to investigate the effects they have on mobility in different scenarios.
The output of the shared mobility simulation in PTV MaaS Modeller consists of individual travel data such as travel distance of the vehicles or waiting time of the travelers. For a more comprehensive analysis of the resulting traffic, e.g., the overall traffic flow on different roads across the service area, the simulation results need to be exported back into PTV Visum and aggregated.
4 Case Study: Investigating the Effects of Shared Mobility in Gothenburg – Mölndal – Partille
4.1 The Gothenburg – Mölndal – Partille (GMP) Area
We apply the presented model to simulate and analyze how the shift from private cars and traditional public transport to carsharing and ridesharing might affect mobility in Gothenburg. The city of Gothenburg has approximately 590 000 inhabitants, making it the second-largest city of Sweden. Every day, approximately 124 000 commuters travel to Gothenburg for work, which also affects the traffic volume during peak hours. Overall, approximately 40% of the daily private traffic consists of work trips. In some suburbs of Gothenburg, more than 70% of the employed inhabitants commute to a workplace, which lies outside of their home municipality. Thus, in this study, we investigate the urban area of Gothenburg (Tätorten Göteborg), which also includes the municipalities of Mölndal (70 000 inhabitants) and Partille (38 000 inhabitants). In total, the Gothenburg–Mölndal–Partille (GMP) region has an area of 200 km2 and almost 700 000 inhabitants.
4.2 Generating Demand for Shared Mobility
In accordance with the GMP, in our study, we model the (origin–destination-based) travel demand in the morning hours of a workday. There are 1.1 million daily trips in the city of Gothenburg and almost 1.6 million trips including the surrounding municipalities [28]. The mode choice in the region is shown in Table 2 considering both inner-city traffic (Göteborgs stad) as well as traffic from the outside (Göteborgsregionen).
Table 2
Mode choice in GMP area based on 2014 travel survey
Mode of trip
Frequency
Car
56.1%
As driver
47.2%
As passenger
8.9%
Walk
19.9%
Public Transport
17.8%
Bike
5.6%
The travel demand is estimated based on a travel survey from 2014 [28]. In this survey, the most common trip purpose is getting to and from work (see Table 3). Other trip purposes include leisure (11%) and shopping (10%). Car trips have an average length of 12 km (20 min) and PT trips 13 km (35 min). In 2010, the number of cars registered in Gothenburg was 150 000, which is 301 cars per 1 000 inhabitants. More than 60% of the households in Gothenburg do not own a car. For the investigated area, there are 90 108 trips by private cars that can be replaced with shared mobility and combined with PT users, a total of 196 643 trips are subject to a shift to shared mobility.1
Table 3
Most common trip purposes. Other purposes include school, visits, and accompanying others on their trips
Purpose of trip
Frequency
Home
38%
Work
19%
Leisure
11%
Daily shopping and services
10%
Others
22%
The zone-to-zone travel demand per mode, which is part of the OD matrix, allows us to simulate the current mobility in Gothenburg before the introduction of shared mobility as a reference scenario for comparing the simulation results for other demand patterns. To simulate the shift to shared mobility, some or all travelers of a specific travel mode will use carsharing or ridesharing for their trips instead of their original travel mode. We further assume that those traveling by bike or walking will not change their travel mode and, thus, not use the shared vehicle service.
In the OD matrices from the GMS that we use as input for the simulations, travel demand is represented as number of trips between zones. In the simulation, each trip is randomly allocated to an access node within the origin zone and an egress node within the destination zone. These access and egress nodes need to be located at a link suitable for walking, i.e., no motorways, bridges, tunnels, or a speed limit of more than 60 km/h. Applying these requirements, we created 16 461 access and egress nodes in the GMP area.
Each shared mobility trip starts and ends at a PUDO. For a location to qualify as PUDO, multiple criteria must be met. This includes that car traffic needs to be allowed on the road and that it cannot have more than 2 lanes in each direction, which we use as an indicator for busy roads with a high traffic volume, e.g., motorways, where it is difficult or prohibited to stop. The maximum speed may not exceed 60 km/h and PUDOs cannot be located in tunnels, on roundabouts, on bridges, or on motorway ramps or exits. Finally, PUDOs should be located where two street segments (links) meet and not at the end of a link, e.g., a dead-end road in a housing area. After identifying all potential locations, we iteratively remove potential locations until a certain minimum distance between the PUDOs is achieved. In total, we were able to generate 4 814 PUDOs with a distance of at least 100 m between the PUDOs (see Fig. 2a). This is, to facilitate the routing of the shared vehicles by limiting the number of PUDOs.
Fig. 2
Distribution of (a) PUDOs and (b) holding areas in GMP area
×
Finally, holding areas are required, which vehicles are allocated to at the start of the simulation and where they wait for trips when not being in service for more than 15 min. For this study, we define 49 holding areas, mostly located in the city center where most demand occurs (see Fig. 2b). Yet, some of the holding areas are located in the suburbs, to reduce trip times for the vehicles. The locations of these holding areas were not chosen to optimize travel distances or walking times but such that all PUDOs can be reached within 10 min at the most.
The remaining parameters for the simulation are chosen as shown in Table 4. Shared mobility might result in increased travel times compared to direct trips. To account for this, travelers will only accept trips that do not exceed 1.5 times the travel time of a direct trip. This detour factor is complemented by an always accepted (10 min) and a maximum accepted (30 min) detour time, as upper and lower boundaries, which corresponds to the travel time ratio between private and public transport currently experienced in medium sized European cities [27]. Implementing push and pull measures can further affect this factor. In 2013, for instance, Gothenburg has implemented a dynamic congestion tax for vehicles entering and exiting central parts of the city at different times of the day. As a result of this, the traffic volume on the major thoroughfares decreased by ca. 12% [29].
Table 4
Parameters of the simulation model
Parameter
Value
Prebooking time
0 min
Acceptable waiting time
10 min
Always accepted detour time
10 min
Maximum detour factor
1.5
Maximum accepted detour time
30 min
Board and alight time per trip request
1 min (each)
Maximum idle time of shared vehicles before relocating to holding area
15 min
Simulated time interval
6 – 10 a.m
Traveler group size
88% single, 10% group of 2, 2% group of 3
Based on the data from the travel survey, we assume the ratio of travelers requesting a trip for only one person to be 88%. In addition, we assume that 10% of the travelers will book a trip for two persons and 2% for three persons. The time window that we simulate is the morning rush hour, between 6 and 10 a.m. Finally, the service is provided with only one type of vehicle that provides space for up to 5 passengers and the boarding as well as alighting time per trip is assumed to be 1 min.
In our study, we do not limit the number of available vehicles. This is to analyze how many vehicles are needed to serve the demand for shared mobility. Yet, due to waiting times and geographical limitations, it is possible that not all demand for shared mobility trips can be met. In this case, we assume the use a private car for the trip.
4.3 Investigated Carshare and Rideshare Scenarios
The study investigates the effects that different service configurations as well as different levels of demand have on both the mobility in the GMP area but also on the quality of service that is experienced by the travelers. To be able to compare the results of the different scenarios, we define a base scenario (baseline), which we use as a reference for the other scenarios. It represents the current mode distribution before shifting to shared mobility. For all scenarios, we simulate the morning peak hours (6–10 a.m.) on a weekday. During this time, there are about 104 000 private car users (drivers and passengers) and about 92 000 public transport users, traveling within GMP.
There are some assumptions that apply for all scenarios. We simulate the use of shared vehicles only within the GPM area and all vehicles have the same capacity of 5 passengers. This means that trips into and out of the GMP area will not use shared vehicles but remain using their current modes of transport. The same network of access and egress nodes is used in all scenarios and shared vehicles will only stop at designated PUDOs. The demand, which is defined by OD matrices, is fixed and will not be affected by changed travel options. Moreover, we only consider private travelers and exclude commercial traffic, e.g., transport of goods, when simulating the shift to shared mobility. Hence, we assume that travel times are independent of the traffic volume. Finally, our simulation does not investigate the gradual introduction of shared mobility but assumes that it is fully established.
To analyze the effects of a potential shift from private car drivers to ridesharing (R1) and carsharing (C1), we construct a scenario for each case, in which we used the most plausible setting of the parameters as also being the default setting (see Table 4). To study the effects of the volume of trips, this scenario is complemented by (R2), in which only \(\frac13\) of the private car travelers switch to ridesharing. Given that we have OD matrices for public transport (PT) users, we also analyzed a scenario (R3), where public transport users are transferred to ridesharing. However, to make them comparable with scenarios concerning private car users, we scaled the demand of PT users to the same demand level as for car users. This allows for investigating differences between PT users and private car users when the shifting to shared mobility without the potential bias of different efficiency levels due to different demand levels. To understand the effects of the most extreme situation, i.e., all current private car users and PT users are shifted to ridesharing, a scenario (R4) representing this situation was created. The scenarios mostly focus on the shift to rideshare and do not simulate the shift from PT to carshare as this seems most unrealistic. Obviously, one can study the effects of different assumptions on the services, yet, we choose to study most scenarios with a 100% shift to the respective mode to get clear indications on the potential effects.
To study potential measures for increasing the ridesharing level, we created scenario (R1A) in which the service quality was reduced by increasing the detour factor to 1.75 (instead of 1.5). It can be assumed that an increased detour acceptance will lead to an increasing occupancy of the vehicles as it facilitates combining trips of travelers. In this scenario, we also increased the always accepted detour time from 10 to 15 min, which mostly affects short trips, and the maximum accepted detour time from 30 to 45 min. Moreover, we also created scenarios (R1B) where we investigate how reducing the number of PUDOs and the resulting greater distance between PUDOs affect mobility. For this purpose, we increased the distance between PUDOs from at least 100 m, which is the default setting, to 300 m (R1B1) and 1000 m (R1B2), reducing the number of PUDOs from 4 729 to 959 and 127 respectively. Finally, in scenario R1B3, we removed all PUDOs (81) from the city center (within Vallgraven, an area of 0.64 km2) to investigate the effects a car-free inner city has. In Fig. 3, an overview of all simulated scenarios is presented.
Fig. 3
Overview of the simulated scenarios. All scenarios except R1A and R1B assume a maximum detour factor of 1.5, a detour time of 10–30 min, and a minimum PUDO distance of 100 m
×
5 Results of the Simulation Study
In this section, we present and compare the results from the different scenarios we simulated. To analyze the results with respect to the questions we want to address, we have chosen to include the following performance indicators. For the vehicles, we measure the total vkm (the cumulative amount of kilometers driven by all vehicles providing shared mobility services both with and without passengers), total number of vehicles (that are required to satisfy the entire demand for shared mobility trips), average vehicle time for shared vehicles (the service time of shared vehicles, i.e., with one or more passengers on board), and the average occupancy of the vehicles when being in service.
From the travelers’ perspective, we measure the average travel time (time spend in a vehicle), average walking time (from access node to origin PUDO and from destination PUDO to egress node), average waiting time (at origin PUDO waiting for vehicle to arrive), and the average total trip time (sum of walking, waiting, boarding, and in-vehicle travel time). An overview of the results is provided in Tables 5 and 6. The presented outputs only concern trips within the GMP area and do only include trips of private persons.
Table 5
Simulation results of different scenarios, where private car and public transports trips are replaced with carshare and rideshare. The results of the baseline scenario are provided as reference and the difference from the baseline scenario is provided in parentheses
All private car trips to
\(\frac13\)of private car trips to
All PT trips (scaled) to
All PT and private car trips to
Private car trips
rideshare (R1)
carshare (C1)
rideshare (R2)
rideshare (R3)
rideshare (R4)
Baseline
Vehicles
Total vkm (in thousands) of private cars, rideshare and carshare
688 (-17%)
931 (13%)
792 (-4%) (-12%*)
610 (-26%)
1 203 (46%)
824
Total number of shared vehicles [rounded to hundreds]
14 900 (-83%)
24 800 (-71%)
5 300* (-81%*)
20 300 (-76%)
34 000 (-60%)
85 500
Average non-empty service time of shared vehicles (hours)
1.18 (462%)
0.85 (305%)
1.15* (379%*)
0.82 (290%)
0.94 (348%)
0.21**
Occupancy (Average number of travelers when on route)
1.97 (54%)
1.28 (0%)
1.48
1.89* (16%)
2.19 (71%)
2.09 (63%)
1.28
Travelers
Average in-vehicle travel time (min.)
15.35 (23%)
12.49 (0%)
15.31* (23%*)
15.36 (23%)
16.11 (29%)
12.49
Average walking time (min.)
3.33
3.33
3.38*
3.37
3.36
-
Average waiting time (min.)
5.51
6.50
5.02*
5.82
4.86
-
Average travel time (walking, waiting, boarding, alighting and in-vehicle time) (min.)
26.19 (110%)
24.32 (95%)
25.71* (106%*)
26.55 (113%)
26.33 (111%)
-
*Concerns only \(\frac13\) of private car travelers switching to rideshare, remaining \(\frac23\) use private car. **Service time of private cars
Table 6
Simulation results of different variations of the scenarios, where all private car trips are replaced with rideshare. In scenario R1A, the detour factor is 1.75 (instead of 1.5) and the accepted minimum and maximum detour time is 15–45 min (instead of 10–30 min). In scenarios R1B, the minimum distance between the PUDOs is changed from at least 100 m to 300 m in R1B1 and to 1000 m in R1B2. In scenario R1B3, there are no PUDOs in the city center (minimum distance between PUDOs is 100 m)
All private car trips to rideshare
(R1)
(R1A) incr. detour acceptance
(R1B1) 300 m between PUDOs
(R1B2) 1 km between PUDOs
(R1B3) no PUDOs citycenter
Vehicles
Total vkm (in thousands) of private cars, rideshare and carshare
688 (-17%)
566 (-31%)
641 (-22%)
563 (-32%)
685 (-17%)
Total number of shared vehicles [rounded to hundreds]
14 900 (-83%)
10 900 (-87%)
14 000 (-84%)
11,600 (-82%)
14 700 (-83%)
Average non-empty service time of shared vehicles (h)
1.18 (462%)
1.47 (600%)
1,.8 (462%)
1.17 (457%)
1.19 (467%)
Occupancy (Average number of travelers when on route)
1.97 (54%)
2.50 (95%)
2.07 (62%)
2.30 (80%)
1.99 (55%)
Travelers
Average in-vehicle travel time (min.)
15.35 (23%)
20.51 (64%)
16.81 (35%)
13.32 (7%)
17.31 (39%)
Average walking time (min.)
3.33
3.33
10.28
35.66
3.88
Average waiting time (min.)
5.51
5.99
5.49
5.86
5.52
Average travel time (walking, waiting, boarding, alighting and in-vehicle time) (min.)
26.19 (110%)
31.83 (155%)
34.58 (177%)
56.84 (355%)
28.71 (130%)
5.1 Comparison of Rideshare and Carshare
The results indicate that the total vkm for people using shared vehicles increases with carsharing (13%) and decrease with ridesharing (-17%) compared to using private cars. Notably, in the default set-up of the service, the reduction in vkm with ridesharing is limited, 17% or less depending on the fraction of private car users that switch. If only one-third of the private car users switch to shared mobility, the reduction is 12% for that group. These relatively low reductions motivate additional experiments on how to potentially increase the level of ridesharing, which are presented below.
The simulation results also clearly underline the potential to reduce the number of vehicles, which are required between 6 and 10 a.m. during a weekday within the GMP area. The relative reduction of vehicles is at a level of 83% (with a complete shift to rideshare) and 71% for the carshare scenario. Hence, a shared self-driving vehicle can replace almost 6 private cars in the default setting of the service during the simulated period.
When studying the relative usage in terms of the vehicles’ service times, we can observe differences between the scenarios. Given that a private car is only used for 0.21 h (i.e., 13 min) during the investigated period, shared mobility solutions operated as carshare or rideshare can increase the utilization of vehicles to slightly more than an hour in the rideshare and slightly less than an hour in the carshare scenario. Notably, the OD demand associated with PT users results in a lower utilization. This is as the number of required vehicles is higher for the PT-user scenario. One explanation is that the flow of PT-users is much more one-way directed on some main commuting routes, e.g., inbound for the city center at the morning, meaning that shared vehicles typically need to go back-and-forth with empty running or as in the simulation by utilizing additional vehicles. This might, for instance, be a result of PT-users not being evenly distributed across the investigated area due to variations in PT accessibility.
The simulation results show some potential effects for the traveler. The average in-vehicle travel time increases by 23% in case all private car drivers shift to rideshare. The average waiting time for being picked up is about 5 to 6 min. Obviously, this number could have been reduced if prebooking was used or if it would be possible to predict where demand occurs. When estimating travel times, changes in traffic volume and potential effects of congestion are not considered. This is because of the uncertainty how shared and autonomous mobility will affect traffic flow.
The results also indicate that there might be a tradeoff for providing carsharing with no extra driving time but with an increased waiting time (6.5 min) compared to ridesharing with less waiting time (5.51 min) but of course with more in-vehicle time. The extra waiting time of carshare is likely due to an increased number of cases with scarcity of nearby cars. Moreover, the measured average walking time is just above 3 min. This observation is, however, difficult to compare to the baseline scenario as we lack data on the average walking time to the parking of private cars.
The simulation tool provides an estimation of the average number of travelers sharing a vehicle, i.e., the number of co-travelers each traveler meets during a trip. Compared to this, current private car users share their vehicle with 1.28 people on average. Due to rideshare, this number increases to around 2, which is not particular impressive compared to what a person would experience on a bus in a city environment during the morning peak.
5.2 Improving Rideshare
To explore potential approaches to increase ridesharing or maybe more importantly to reduce vehicle kilometers, we also conducted experiments where the service is less attractive. This includes configurations with a higher detour factor and higher allowed detour time as well as experiments with greater distance between the PUDOs and removing PUDOs from the city center (see Table 6).
It can be observed that both investigated changes (increased detour factor and PUDO distance) result in a reduction in vkm compared to the original configuration of the scenario (R1). Increasing the minimum distance between PUDOs from 100 to 300 m and 1000 m reduces the number of required PUDOs by around 80% each (decrease from 4 729 PUDOs to 959 and 127 respectively). Furthermore, a distance of 300 m reduces vkm by an additional 5% and by 15% in case of 1000 m distance. The number of required vehicles can be reduced by more than 20% compared to the 100 m scenario and the occupancy increases from 1.97 to up to 2.3.
Increasing the distance between PUDOs also affects travel times. Increasing from 100 to 300 m increases the average total travel time by 36% from 26.19 to 34.58 min. This is not only due to increased walking time to the PUDO but also due to the in-vehicle travel time increasing by almost 10%. A 1000 m distance, however, reduces the in-vehicle travel time to 13.32 min (-13%). Yet, the total travel time increases to 56.84 min, which is more than twice the travel time of a PUDO distance of 100 m.
Finally, in the scenario where all PUDOs were removed from the city center (R1B3), the total vmk (685; -0,5%) as well as the number of required shared vehicles (14 700; -1,5%) are slightly lower compared to scenario R1, where PUDOs also exist in the city center. Total travel time, however, increases by 2.5 min, which is due to increasing in-vehicle travel time (+ 13%) and walking time (+ 17%).
In case of a changed detour factor, the number of vehicles could also be significantly reduced. The in-vehicle travel time increases with the changed detour factors and the reduction of the PUDOs leads to a slightly reduced in-vehicle travel time while, of course, the walking time increases.
5.3 Effects on Different Road Types and City Areas
Many studies simulating and investigating how shared transportation solutions affect mobility provide global analyses, i.e., measure changes in vkm for the entire city. We study the changes in traffic volume to investigate whether shared mobility has different effects on different areas of the city. Figure 4a illustrates the changes in traffic flow for scenario R1, where all private car trips are replaced by rideshare, whereas Fig. 4b shows the changes in traffic flow for scenario C1, where all private car trips are replaced by carshare, to compare the effects rideshare and carshare have on traffic volume. Green links represent a reduction of traffic flow compared to the base scenario whereas orange and red lines represent an increase. A reduction of traffic can be observed in the city center while an increase can be observed on smaller streets such as residential areas. Note that the total traffic in these scenarios includes traffic in and out of the areas as well as commercial vehicles and goods vehicles.
Fig. 4
Illustrating the change in traffic flow (number of vehicles) of (a) R1 (rideshare) and (b) C1 (carshare) compared to Baseline; green representing a reduction and red an increase
×
Changes in traffic volume are not only due to the simulated scenario but do also differ between the road types (see Table 7). We investigate the changes in traffic volume on four types of roads: arterial roads (thoroughfares leading into or through the city), river crossings (traffic bottlenecks), central roads (major roads within city), and residential areas (without through traffic). The shift to shared mobility has the least effect on through roads, which face high traffic volumes during rush hours. In scenario R1, a reduction in traffic volume generated by the investigated groups of travelers of -17% can be achieved. In contrast, the traffic volume in residential areas increases significantly in all scenarios.
Table 7
Change in traffic volume for different road types. All values are relative to the baseline scenario without shared mobility
Roadtype
R1
C1
Arterial road
-17%
1%
River crossings
-12%
0%
Central roads
-3%
21%
Residential areas
4%
29%
6 Discussion of Results
Based on the simulated scenarios, we have gained some new insights on the effects of shared mobility and confirmed some findings from earlier related studies. A comparison of the results of the related studies and of our study is presented in Table 8, which focuses on the scenario where all car users shift to rideshare.
Table 8
Results of the related studies for the scenario, where all private car users shift to rideshare. The studies from Stuttgart and Lyon were not comparable due to the simulated scenarios. Average total trip duration includes walking, waiting, boarding and alight as well as the travel time in vehicle
Number of vehicles
Passenger kilometers ridesharing (in mio)
Waiting time (min)
In-vehicle travel time (min)
Total trip duration (min)
Mean occupancy
Lisbon
21 120
4 010 (6%)
3.8
15.9
19.7
-
Helsinki
20 522
16 320 (-33%)
9.6
17.9
-
2.32**
Auckland
44 553
62 200 (-51%)
2.0
-
28.9
2.34**
Dublin
26 538
33 454 (-42%)
-
-
23.0
2.3**
Oslo
26 000
3 700* (-14%)
2.9
17.7
20.5
1.62
Gothenburg
14 900
688* (-17%)
5.5
15.4
26.2
1.97
*vehicle kilometers **shared taxi with seats for 8 passengers
6.1 Traffic Volume
In terms of the traffic volume, we observed that the transition from private cars to ridesharing can lead to a significant reduction (-17%) in vkm in this group of travelers. This effect, however, could not be observed for the shift to carsharing, where the traffic volume increased compared to the baseline scenario (13%). These results correspond to those of the related studies, which confirm that the most noticeable reduction in vkm can be expected when all car drivers shift to ridesharing (Oslo: -14%, Helsinki: -33%, Dublin: -42%).
In our simulations, the decrease in vkm that can be achieved by ridesharing depends on the efficiency of the fleet managers, the service offered, the number of travelers, and the occupation of the shared vehicles. The results clearly indicate a potential of reducing the traffic volume with shared mobility services. Notably, the average vehicle occupancy, measured over the entire time of the ride, is not particular impressive (1.97). This is far from current PT-services in the GMP area during the considered time period. Yet, similar numbers were observed by the related studies, e.g., the Oslo study, where the mean occupancy in operation for ridesharing is 1.86 compared to 1.14 of private cars in the baseline scenario. In the Dublin study, an average occupancy of 2.3 passengers was estimated for shared taxis. One potential explanation is that the simulator does not manage to optimize the service for ridesharing particularly well. Another explanation is that it is rather difficult to achieve a similarly high ridesharing level as for a PT-service since travelers are not gathered at interchange points (which is the case for some PT-travelers).
When shifting both private car trips and PT trips to shared mobility, only a minor increase in occupancy can be observed (2.06). In this case, it must be noted that the trips are assumed to start or end from PUDOs close by to true origin and destinations, not from original PT stops.
Compared to private car trips, the shift to shared mobility might be assumed to lead to an increased traffic volume as the shared vehicles need to be relocated between trips. Instead of parking the vehicles at the destination of their previous trip while waiting for the next trip, they are redirected to holding areas. This reduces the congestion along the streets due to less parked vehicles and the strategic positioning might lead to reduced waiting times for passengers. Yet, this also leads to approximately 12% of the vkm being empty repositioning trips. Similar values between 15 and 20% were observed in the related studies.
The overall reduction in vkm and the reduced congestion during peak hours that can be observed for some scenarios often results from a clear reduction in traffic on the major traffic routes. Yet, the need to reposition the shared vehicles after a trip or before results in increased traffic in some of the outer parts of the network. This observation was confirmed by the related studies, e.g., in Helsinki, where a traffic relief in the city center was accompanied by increased congestion on secondary roads which lead to, for instance, park and ride facilities. Also in Lisbon, an increased utilization on local road networks with close connection to PT was observed. The explanation, we believe, is mainly due to a potential higher ridesharing level on the main routes, whereas the ridesharing is lower, and the empty running is higher in the outer parts since people are picked at their homes in the outer parts.
The conclusions we draw regarding how shared mobility affects congestions are based on vkm, number of vehicles, and the resulting traffic density. The simulation model itself does not explicitly consider congestion and travel times are not affected by traffic volume. This is because of the uncertainty regarding how autonomous vehicles will affect traffic flow and how the future road network will look like. Still, the base scenario implicitly considers congestion due to RVU data that is part of the GSM and used for generating the OD matrices, which serve as simulation input data. Hence, we can assume that our results underestimate the presumably positive effects autonomous mobility has on congestion and travel times.
6.2 Number of Vehicles
A major advantage of shared mobility is the significant reduction in number of vehicles that are required. The results of our simulations indicate that there is a potential that shared vehicles can replace at least 5 private cars during the morning peak in case of ridesharing and almost 4 vehicles in carsharing scenarios (see Table 5). Already the shift of one-third of today’s private car trips to ridesharing would make the driving and parking of 28 000 private cars obsolete, however, it results in a need of 4 200 shared vehicles (reduction of 83%), which require temporary stops and holding areas. Overall, replacing all private car trips with ridesharing could be achieved by using 17% of today’s vehicles. In the related studies, reductions as low as only 2% to 10% of the baseline were reported. Yet, some of the studies simulated the mobility of an entire day and not just the morning peak hours.
In our study, 90 108 trips of private car users were served by 14 806 shared vehicles, which corresponds to 6 trips per vehicle. In case PT users also shift to shared mobility, a similar utilization can be observed.
During the morning peak, 15.4 trips per vehicle were served in Olso and 31.4 in Dublin. The ratio of PT travelers in Dublin is lower compared to the other cities, which increases the demand for shared mobility when removing private cars. Moreover, vehicles of different size are used to transport up to 16 travelers at a time, which increases the vehicles’ occupation. In the other studies, the average number of daily trips being served using a single shared vehicle varies between 34 and 66.
6.3 Quality of Service for Travelers
From a travelers’ perspective, the shift to shared mobility is associated with an increased travel time especially compared to using a private car. This increase results from waiting times as well as detour and boarding times due to other travelers being picked up or dropped off. In case all private car trips shift to ridesharing, we estimate the average waiting time to be 5.5 min in addition to a walking time of 3.3 min, leading to a total average travel time of 26.19 min. This corresponds to an experienced detour factor of 1.7, i.e., the entire trip using shared mobility including walking and waiting is 1.7 times longer than a direct trip. Comparing the direct trip travel time of 15.4 min to the simulated in-vehicle travel time of 17.4 min, the increase of travel times with ridesharing is less than 15% on average considering the detour and pick-up times of other travelers. The waiting time can be reduced if trips are pre-booked or good prognosis are made to have vehicles available closer to expected demand. Related studies showed experienced detour factors between 1.53 and 1.87 with average waiting times between 3 and 4 min.
6.4 Trade-Offs
The results of the study indicate that the total vkm, the number of required vehicles, and the level of service are mostly affected by the design of the service (e.g., the distance between PUDOs or the size of the vehicles) as well as of the number of passengers who use the service and their detour acceptance. The feasibility and positive effects of shifting from private cars to rideshare could be shown. Yet, the study also revealed the limitations of carsharing, and challenges associated with the shift from PT to shared mobility.
7 Conclusions
This article presents the results of a simulation study on the massive implementation and usage of shared vehicles in the GMP area (Gothenburg-Mölndal-Partille) in Sweden. The study comprises a what-if analysis to investigate the potential effects of a shift to shared mobility assuming that respective push and pull measures were successfully implemented. We simulate the shift from private cars and public transport to carsharing and ridesharing for journeys within GMP, yet, ignoring potential shared vehicles to and from the GMP area and trips combining shared mobility and traditional PT.
Based on the results of our study, we can summarize the following conclusions:
A shift from private cars to carshare results in 35% higher vkm compared to rideshare.
Replacing all private car trips with rideshare will reduce the vkm by up to 17% in contrast to carshare, which increases the vkm by up to 13%.
The average vehicle occupancy when private car trips are replaced with rideshare is 1.97.
Up to 12% of the vkm generated by shared mobility are empty repositioning trips.
Shared vehicles can replace at least 5 private cars during the morning peak in case of ridesharing and almost 4 vehicles in carsharing scenarios.
An increased detour acceptance can significantly reduce the number of required vehicles and vkm as well as increase the occupancy of the shared vehicles.
A greater distance between PUDOs decreases vkm, however, also significantly increases the average total travel time.
Our simulations showed that shared mobility has the potential of reducing traffic volume. However, the reduction is greatest on arterial roads and in residential areas the volume might increase. Even though both types of services (rideshare and carshare) might coexist, a strong focus on rideshare and a reduction of carshare is required to achieve a reduction in vkm. Even though a shift from public transport to shared mobility generate slightly less vkm of rideshare vehicles than an equivalent shift from private cars, it is important that the major public transport lines should be retained to avoid a dramatic increase of traffic volume due to a potential shift away from public transport.
In our simulation model, we assume that there is only one single shared mobility service provider. A greater demand facilitates the routing and scheduling of the vehicles and leads to reduced waiting times due to the larger number of available vehicles. In the future, however, it can be expected that there will be a variety of service providers, which might affect the convenience and efficiency of the service. It needs to be investigated how the existence of different providers affects both the traffic volume as well as the quality of service for the passengers. This might lead to new insights regarding required cooperation between the providers or the need for policies to cope with the resulting effects.
The simulations also indicated a relatively low rideshare level (occupancy) compared to the size of the vehicles. With respect to the efficiency and ultimately also the (economic) viability of shared mobility services, however, the rideshare level might need to increase. We identified detour tolerance as one reason for low occupancy of the vehicles, which might be overcome by optimizing the route planning. Another reason for long travel times is the waiting time for the vehicles to arrive, which might be decreased by the prebooking of trips. This, however, might require more advanced planning and replanning of trips. Finally, there is a need to investigate other first- and last-mile solutions or the interchange between ridesharing vehicles as measures to increase vehicle occupancy and travel times.
Promoting and governing the shift to shared mobility requires policy changes and the implementation of suitable push and pull measures that discourage the use of private vehicles and attract travelers to ridesharing solutions. Hence, to expedite this shift, it is important for authorities and policymakers to continue developing appropriate measures that facilitate shared mobility and to continue investigating how shared mobility can become a part of a sustainable society.
Acknowledgements
We would like to thank all members of the Eldsjäl project, in which this study has been carried out. This research is supported by Vinnova through the Drive Sweden strategic innovation program under grant number 2019-05094, Västra Götalandsregionen under grant number KTN 2019-00124, K2 (The Swedish Knowledge Center on Public Transport), and the Internet of Things and People Research Center at Malmö University. This work was partially supported by the Wallenberg AI, Autonomous Systems and Software Program – Humanities and Society (WASP-HS) funded by the Marianne and Marcus Wallenberg Foundation.
Declarations
Conflict of Interest
The authors declare that they have no conflict of interest.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
ATZlectronics worldwide is up-to-speed on new trends and developments in automotive electronics on a scientific level with a high depth of information.
Order your 30-days-trial for free and without any commitment.
Die Fachzeitschrift ATZelektronik bietet für Entwickler und Entscheider in der Automobil- und Zulieferindustrie qualitativ hochwertige und fundierte Informationen aus dem gesamten Spektrum der Pkw- und Nutzfahrzeug-Elektronik.
Lassen Sie sich jetzt unverbindlich 2 kostenlose Ausgabe zusenden.
Die im Laufe eines Jahres in der „adhäsion“ veröffentlichten Marktübersichten helfen Anwendern verschiedenster Branchen, sich einen gezielten Überblick über Lieferantenangebote zu verschaffen.
