Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com
ScienceDirect
ScienceDirect
Available online at www.sciencedirect.com
Transportation Research Procedia 00 (2022) 000–000
Transportation Research Procedia 00 (2022) 000–000 www.elsevier.com/locate/procedia
ScienceDirect www.elsevier.com/locate/procedia
Transportation Research Procedia 72 (2023) 3972–3978

Transport Research Arena (TRA) Conference

Transport Research Arena (TRA) Conference
Optimization of DRT concepts using the SUMO simulation
Optimization of DRT concepts using the SUMO simulation
software: How can the service quality be improved?
software: How can the service quality be improved?
Evnika Davida, Filiz Kurta,*, Benedikt Scheiera, Saskia Mittaga
a
Evnika Davida, Filiz Kurta,*, Benedikt Scheiera, Saskia Mittaga
Deutsches Zentrum für Luft- und Raumfahrt e.V. (DLR), Lilienthalplatz 7, 38108 Braunschweig, Germany
a
Deutsches Zentrum für Luft- und Raumfahrt e.V. (DLR), Lilienthalplatz 7, 38108 Braunschweig, Germany

Abstract
Abstract
Within the Reallabor project in Hamburg future digital mobility is being tested with the objective to enhance mobility and transport
in a metropolitan
Within the Reallaborregion suchinasHamburg
project Hamburgfuture
towards environmental
digital and climate
mobility is being friendly
tested with mobility.toToenhance
the objective fulfil this aim isand
mobility necessary to
transport
in a only
not metropolitan
focus onregion such as
the central Hamburg
city, but alsotowards
on the environmental
urban districts.and climate
Hence, withfriendly
the helpmobility. To fulfil thissimulation
of the microscopic aim is necessary to
softw are
SUMO
not onlyitfocus
was analysed how autonomous
on the central on-demand
city, but also vehicles
on the urban can beHence,
districts. used inwith
suburbs like of
the help Bergedorf to improve
the microscopic the connection
simulation of
softw are
such
SUMO suburban areas to the
it was analysed howexisting publicon-demand
autonomous transport system. The
vehicles canaimbeof thein
used research
suburbswasliketoBergedorf
find out what quality the
to improve of service can be
connection of
achieved with on-demand
such suburban areas to the transport (DRT)transport
existing public in Bergedorf andThe
system. howaimit can be optimised
of the research wasin terms
to findofout
users.
whatThe paperofwill
quality present
service canthe
be
achieved
results of with
the simulation
on-demandand transport
the analysis
(DRT)ofintheBergedorf
results. and how it can be optimised in terms of users. The paper will present the
results
© 2022 of The
the simulation and the analysis
Authors. Published of the results.
by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license
© 2023 The Authors. Published by ELSEVIER B.V.
© 2022 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license
(https://creativecommons.org/licenses/by-nc-nd/4.0)
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility
responsibility ofof the
the scientific
scientific committee
(https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under committee of
of the
the Transport
Transport Research
Research Arena
Arena (TRA)
(TRA)Conference
Conference
Peer-review
Keywords: under responsibility
Demand-Responsive of the Simulation;
Transport; scientific committee of the Transport
SUMO; Connected Research
and automated Arenamobility;
multimodal (TRA) sharing
Conference
and pooling; urban
planning.
Keywords: Demand-Responsive Transport; Simulation; SUMO; Connected and automated multimodal mobility; sharing and pooling; urban
planning.

1. Introduction
1. Introduction
The project Reallab HH aims to transform mobility and public transport in a metropolitan region such as Hamburg
The project
towards Reallab HHand
environmentally aims to transform mobility
climate-friendly mobilityand
bypublic transport
developing in a metropolitan
recommendations forregion
actionsuch as Hamburg
(Schlenther et al
towards environmentally
2022). The study examines, andamong
climate-friendly
other things,mobility by developing
how autonomous recommendations
on-demand vehicles canforbeaction (Schlenther
used in et al
urban districts
2022).
in orderThe
to study
improve examines, among other
the integration things,
of such areashow autonomous
with the existingon-demand vehiclessystem.
public transport can be For
usedthis
in urban districts
reason, a test
in order toofimprove
operation the integration
three driverless on-demand of such areas
shuttles with provided
(DRT) the existing public transport
by EasyMile GmbH wassystem. For thisinreason,
introduced autumna2021
test
operation of first
to cover the threeand
driverless on-demand
last mile shuttles
of passengers (DRT)to
travelling provided
and frombythe
EasyMile
S-BahnGmbH
stationwas introducedan
in Bergedorf, in urban
autumn 2021
district
to
of cover the first
Hamburg (VHH and2020).
last mile of passengers travelling to and from the S-Bahn station in Bergedorf, an urban district
of Hamburg (VHH 2020).

* Corresponding author. Tel.: +49-531-295-1092; fax: +49-531-295-3402.

* E-mail [email protected]
address:author.
Corresponding Tel.: +49-531-295-1092; fax: +49-531-295-3402.
E-mail address: [email protected]
2352-1465 © 2022 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license
(https://creativecommons.org/licenses/by-nc-nd/4.0)
2352-1465 © 2022 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the scientific committee of the Transport Research Arena (TRA) Conference
(https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of the Transport Research Arena (TRA) Conference

2352-1465 © 2023 The Authors. Published by ELSEVIER B.V.

This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of the Transport Research Arena (TRA) Conference
10.1016/j.trpro.2023.11.481
 Evnika David et al. / Transportation Research Procedia 72 (2023) 3972–3978 3973
2 Evnika David et al. / Transportation Research Procedia 00 (2022) 000–000

In order to identify operational and economic DRT service optimisation potential, microscopic simulations with
the microscopic traffic simulation software SUMO (Simulation of Urban Mobility) were used to model the test
operation (Lopez et al. 2018). The objective of the study was to identify under which conditions the service quality of
a feeder service can be increased in urban areas. For this purpose, four Use Cases representing different components
of a DRT service were defined, simulated and evaluated. As a result, the simulated Use Cases were varied and
evaluated on the basis of determined key performance indicators (KPI), such as the percentage of rejected requests,
the pre-booking time, the travel time (ride + walk), average ride length in DRT, the total vehicle-kilometres as well
as the percentage of empty trips and the occupancy rate.

2. Methodology and Use Cases

In order to simulate the test operation in Bergedorf, a four-step approach was chosen: In the first step, the Bergedorf
area was analysed with regards to various factors. For example, the urban structure (residential areas, industrial areas,
etc.), the infrastructure as well as the traffic in the area of Bergedorf based on counting points were researched. In the
second step, the demand for a DRT service in Bergedorf was determined. Here, real passenger figures of the Bergedorf
S-Bahn station based on departures of the S-Bahn lines S2 and S21 as well as the number of passengers arriving and
departing from the S-Bahn station were used to determine the demand. In the third step, it was determined what the
capacity and speed of the DRT vehicles to be deployed in Bergedorf should be. Based on a benchmarking study, a
capacity of 6 seats in autonomous DRT shuttles with a maximum allowed speed of 18km/h was assumed. Furthermore,
it was assumed that the DRT vehicles would operate during peak hours (3-6pm) each with 50 min of service and a
ten-minute break at the S-Bahn station. As a last step, the test operation of the DRT system in Bergedorf was simulated
using the SUMO simulation software (Armellini 2021), see Fig 1. Special attention was paid to the current public
transport system in Bergedorf, i.e. already existing bus stops and possible locations for additional virtual bus stops
(green points). Lastly, potential routes for the DRT service in Bergedorf were defined.

Fig 1. Screenshot of the simulation of a DRT service in Bergedorf as part of Reallabor Hamburg

To determine an optimised DRT service through the simulation, four Use Cases were defined, each with different
standard parameter (in the following called: reference scenarios). The reference scenarios represented different
components of a DRT service in Bergedorf and were differentiated by following standard parameters, which were
used as input variables: Demand of potential passengers; Observation period and area of the DRT service; Fleet size,
speed and capacity of DRT vehicles; “Direct Route Factor”, which sets the average deviation from a direct route
between a passenger's boarding and destination points for the DRT to accommodate additional passengers; Waiting
3974 Evnika David et al. / Transportation Research Procedia 72 (2023) 3972–3978
Evnika David et al. / Transportation Research Procedia 00 (2022) 000–000 3

time for the DRT system to pick up the request. The specific attributes of the reference scenarios of the respective Use
Cases are shown in Error! Reference source not found..

Fig. 2. Overview of the evaluated Use Cases and their respective reference scenarios within Reallabor Hamburg

After generating the simulations of the reference scenarios for each use case, different components of the DRT
service were modified by adjusting the input parameters and simulated, resulting in the so-called scenarios of the Use
Cases. The DRT service of the scenarios of the respective Use Cases were then assessed by means of several KPIs, as
presented and described in the following Table 1. The KPIs refer to the study area during the observation period.

Table 1. Selected KPIs for the Assessment of the DRT Service in Bergedorf within Reallabor Hamburg
KPI Unit Description
Ratio of rejected passenger [-] Ratio of passengers who cannot be served within the study area during
requests the observation period
Walking distance [m] Average walking distance of passengers from the starting point to the
S-Bahn station
Average pre-booking time [min] Duration between the booking of the journey and the actual pick-up by
the DRT vehicle
Average ride length in DRT [km] Ride of the journey of the passenger in a DRT vehicle
Journey Time [min] Duration of the journey for the passenger from the starting to the
destination point, including ride with DRT and any walking routes
Total Vehicle Kilometres [km] Total distance driven by all DRT vehicles in operation
Total occupied Vehicle [km] Total distance driven by occupied DRT vehicles in operation
Kilometres
Total empty Vehicle [km] Total distance driven by unoccupied DRT vehicles in operation
Kilometres
Occupancy rate [-] Utilisation of DRT vehicles in terms of passenger kilometres per
vehicle kilometre
 Evnika David et al. / Transportation Research Procedia 72 (2023) 3972–3978 3975
4 Evnika David et al. / Transportation Research Procedia 00 (2022) 000–000

The reference scenario of Use Case 1 represents a DRT service with a demand of 72 passengers in the central
residential area of Bergedorf. Further, the reference scenario of Use Case 2 describes a DRT service with an extended
demand of 440 passengers in the central residential area of Bergedorf. In the reference scenario of Use Case 3, a
passenger demand of 1426 is examined for the entire district of Bergedorf. All three of the described Use Cases are
carried out during peak hours from 3pm to 6pm. Finally, the reference scenario of Use Case 4 represents the simulation
of a full working day (0-24 h) with an extended passenger demand of 440 passengers in the central residential area in
Bergedorf. In all described Use Cases, a Direct Route Factor of 1.5 and a waiting time for the DRT system to pick up
a request of 300s (= 5min) were assumed.

3. Scenarios and Results

In this paper, the two Use Cases 1 and 2 with their individual scenarios will be presented. Starting with a no-case
scenario, it represents a situation in which no DRT shuttles are used in Bergedorf, as it is currently the case. For this
reason, the average walking distance was considered that potential passengers would have to walk if they started at
home and walked to the S-Bahn station. An average walking distance of 1.15km was determined, based on a passenger
demand of 72 potential passengers.
As for Use Case 1, a demand of 72 passengers, a maximum pre-booking time of 5 minutes, a DRF of 1.5 and 2
autonomous DRT shuttles that can travel at a speed of 18km/h were assumed in the reference scenario. Based on this
reference scenario, four further scenarios were modelled in Use Case 1, which are referred to below as Scenario 1a,
1b, 1c and 1d and are varied with a parameter variation as follows:

a) DRT shuttles run at permissible speed (30km/h minor roads and 50km/h major roads));
b) More DRT shuttles will be used (6 instead of 2 vehicles);
c) Combination of a) and b) → Six DRT shuttles will be used and they will travel at the local speed;
d) More DRT shuttles will be used (6 instead of 2 vehicles with a speed of 18 km/h) and they will travel with
an increase of maximum allowed Direct-Route Factor (10 instead of 1,5).

The following Table 2 presents the results of the simulated scenarios of Use Case 1 with regards to the described
KPIs. The results show that the reference scenario in Use Case 1 is not good enough for operational use, as 34% of
the trip requests could not be served. This leads to lower absolute vehicle kilometres than in the other scenarios. This
is due to the fact that there are not enough vehicles, their speeds are too low or direct route factor is too low. In
comparison, the rejection rates for scenarios 1a-d turned out to be low.
In scenario 1a), the speed of the vehicles is increased to the permitted road speed. This measure leads to
significantly lower journey times (-25%) and pre-booking times (-43%). However, due to the higher speeds, the
vehicle requirements and thus the costs are higher.
In scenario 1b), lower pre-booking times and fewer detours can be achieved through a higher number of vehicles.
The latter can be seen in the travel distance of the passengers, which decreases by 7%. The journey times have also
decreased compared to the reference scenario, but higher than in scenario 1a), in which the vehicles operate at higher
speeds. A combination of the previous two scenarios with higher vehicle numbers and speed is Scenario 1c). This
represents the best quality of service for passengers, as the pre-booking time and journey times are the lowest.
In scenario 1d), in addition to a higher number of vehicles, the maximum diversions allowed (high DRF) to the
scheduling algorithm was considered. It is interesting that the vehicle kilometres driven can be significantly reduced
(approx. 10% compared to scenario 1b with a high number of vehicles). This is achieved by bundling more trips.
Unfortunately, this is in conflict with the quality of service, i.e. the pre-booking time and the journey time increase
(5% and 24% respectively compared to scenario 1b).
It is interesting that the vehicle kilometres driven can be significantly reduced (approx. 10% compared to scenario
1b with a high number of vehicles). This is achieved by bundling more trips. Unfortunately, this is in conflict with the
quality of service, i.e. the pre-booking time and the journey time increase (5% and 24% respectively compared to
scenario 1b).
3976 Evnika David et al. / Transportation Research Procedia 72 (2023) 3972–3978
Evnika David et al. / Transportation Research Procedia 00 (2022) 000–000 5

Table 2. Results of Use Case 1 (low demand)

Use Case Unit Use Case 1

1c) 1d)
reference 1a) 1b)
Scenarios Speed & Vehicles &
scenario Speed Vehicles
Vehicles DRF
Input-Parameter
Direct Route Factor 1,5 1,5 1,5 1,5 10
Speed of DRT vehicles permissible permissible
km/h 18 18 18
speed speed
Number of DRT vehicles 2 2 6 6 6
Output-KPIs
Share of rejected passenger
% 34 6 8 4 8
requests
Walking distance of the
km 0,04 0,04 0,04 0,04 0,04
passenger
Pre-booking time min 8,25 4,71 5,41 3,80 5,67
Ride length in DRT km 1,39 1,59 1,29 1,48 1,60
Journey Time (Walk & Ride) min 6,79 5,12 6,37 4,77 7,91
Total Vehicle-Km km 65,65 96,90 100,70 98,02 89,91
Total occupied Vehicle-Km km 44,31 69,23 69,32 70,66 67,46
Total distance of empty runs km 21,34 27,68 31,39 27,37 22,45
Vehicle Occupancy
0,96 1,06 0,81 0,99 1,13
(passenger-km/vehicle-km)

The utilisation of the vehicles is low with the assumed demand in Use Case 1 (Vehicle Occupancy approx. 0.8 to
1.12). Therefore, scenarios with a higher demand were investigated with Use Case 2.
For Use Case 2, an increased demand with 440 potential passengers was taken as a basis. Here, 13 DRT shuttles
were used in Use Case 2 - reference scenario, each travelling at permissible speed. Also, a maximum pre-booking
time of 5 min and a DRF of 1.5 were assumed. Besides the reference scenario, three further scenarios were analysed
in Use Case 2, in the following referred to as Scenario 2a, 2b and 2c, considering adjusted parameters as follows:

a) DRT service with an extended max. pre-booking time of 30 min (instead of 5 min);
b) DRT shuttles will travel with an increased Direct Route Factor (20 instead of 1,5);
c) More DRT shuttles will be used (27 instead of 13 vehicles).

The reference scenario in Use Case 2 already considers higher vehicle speeds. Therefore, the rejection of trip
requests is in the satisfactory range. The research question was what can be achieved with changed parameters in
relation to the KPI? The following Table 3 presents the results of the simulated scenarios of Use Case 2 with regards
to the described KPIs.
In scenario 2a), by increasing the maximum permitted pre-booking time, it can be achieved that journeys are
bundled more optimally, even if the travellers have to accept higher pre-booking times. However, vehicle kilometres
decrease by about 4% and journey times are also slightly lower (- 1.2%).
In scenario 2b) the direct route factor was increased. This leads to a significantly stronger bundling of journeys,
but also to an increase in journey times of 14%. In this case, the vehicle operators profit from lower vehicle kilometres
by approx. 8.5%.
 Evnika David et al. / Transportation Research Procedia 72 (2023) 3972–3978 3977
6 Evnika David et al. / Transportation Research Procedia 00 (2022) 000–000

In scenario 2c) a further increase in the number of vehicles was analysed. As expected, the pre-booking time
(-9%) and the journey times for travellers (- 3%) decrease. The vehicle kilometres slightly increase.

Table 3. Results of Use Case 2 (high demand)

Use Case Unit Use Case 2

2a)
reference 2b) 2c)
Parameter Pre-booking
scenario DRF Vehicle
time
Input-Parameter
Waiting time for DRT system
min 5 30 5 5
to pick up the request
Direct Route Factor 1,5 1,5 20 1,5
Number of DRT vehicles 13 13 13 27
Output-KPIs
Share of rejected passenger
% 6 9 6 8
requests
Walking distance of the
km 0,04 0,04 0,04 0,04
passenger
Pre-booking time min 3,92 4,47 4,07 3,57
Route length of DRT km 1,69 1,67 1,92 1,64
Journey Time (Walk & Ride) min 5,93 5,79 6,78 5,73
Total Vehicle-Km km 392,72 376,22 359,08 394,37
Total occupied Vehicle-Km km 296,34 283,94 298,21 302,43
Total distance of empty runs km 96,38 92,28 60,87 91,94
Vehicle Occupancy
1,78 1,77 2,20 1,69
(passenger-km/vehicle-km)

4. Conclusion

By the variation of the four Use Cases, it was possible to analyse different DRT services in Bergedorf based on
defined standard parameter, e.g. the demand of potential passengers. Using the SUMO simulation software, the
standard parameters of the reference scenarios have been modified in order to generate further scenarios of the
respective Use Case and to examine their effects on the KPI described above. Hence, the two presented Use Cases 1
and 2 reflected a number of various DRT services, allowing to identify an optimised DRT service in Bergedorf. In
conclusion, in Use Case 1 the increase of speed showed the clearest positive effect on the KPI and should be achieved
by the operators of the DRT-vehicles. Although not the subject of this paper, the cost recovery ratio was calculated
using the net present value method and life cycle costing over 10 years of operation. The cost recovery ratio in this
scenario is about 80%. In contrast, in the scenario with the increased number of vehicles of 6, a cost recovery ratio of
only 35% is achieved. Next to speed, the direct route factor had the greatest impact on increasing vehicle utilisation.
In Use Case 2, the higher demand was clearly leading to better vehicle utilisation (Vehicle Occupancy 1.7 to 2.2),
which leads to a full cost recovery ratio. Thereby, the quality of the service hardly decreases. It can be seen that a
DRT shuttle as station feeder can achieve a good cost recovery ratio together with a good service-quality.
3978 Evnika David et al. / Transportation Research Procedia 72 (2023) 3972–3978
Evnika David et al. / Transportation Research Procedia 00 (2022) 000–000 7

Acknowledgement

This work was funded by a future grant from the German Federal Ministry of Transport and Digital Infrastructure.

References

T. Schlenther, P. Wagner, G. Rybczak, K. Nagel, L. Bieker-Walz und M. Ortgiese, “Simulation-based investigation of transport scenarios for
Hamburg”, 2022. [Online]. Available:
https://reader.elsevier.com/reader/sd/pii/S1877050922004896?token=8FA6F5AFA4022D221AA3123FE98F3E840EA82616630352ABE0A
DB73AD88891AF8DCB6B46922902CF4994448AD3E52F0E&originRegion=eu-west-1&originCreation=20220513065528. [Accessed on 12
May 2022].
VHH (Verkehrsbetriebe Hamburg-Holstein GmbH), “Autonome Shuttles für Bergedorf,”, 17 November 2020. [Online]. Available:
https://vhhbus.de/autonome-shuttles-bergedorf/. [Accessed on 05 May 2022].
P. A. Lopez, M. Behrisch, L. Bieker-Walz, J. Erdmann, Y.-P. Flotteröd, R. Hilbrich, L. Lücken, J. Rummel, P. Wagner und E. Wießner,
“Microscopic Traffic Simulation using SUMO”, 07 November 2018. [Online]. Available: https://elib.dlr.de/127994/1/08569938.pdf. [Accessed
on 05 May 2022].
M. G. Armellini, “Simulation of Demand Responsive Transport using a dynamic scheduling tool with SUMO”, 2021. [Online]. Available:
https://elib.dlr.de/143995/1/sumo2021.pdf. [Accessed on 05 May 2022].