Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com
Available online at www.sciencedirect.com

ScienceDirect
Procedia Computer Science 00 (2019) 000–000
Procedia
Procedia Computer
Computer Science
Science 17000 (2019)
(2020) 000–000
757–763 www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia

The 9th International Workshop on Agent-based Mobility, Traffic and Transportation Models,
The 9th International (ABMTRANS)
Workshop on Agent-based
April 6 - 9,Mobility, TrafficPoland
2020, Warsaw, and Transportation Models,
(ABMTRANS) April 6 - 9, 2020, Warsaw, Poland
Electrification
Electrification of
of Urban
Urban Freight
Freight Transport
Transport –
– aa Case
Case Study
Study of
of the
the
Food Retailing Industry
Food Retailing Industry
Kai Martins-Turnera,∗, Alexander Grahleb , Kai Nagela , Dietmar Göhlichb
a Technische
Kai Martins-Turnera,∗, Alexander Grahleb , Kai Nagela , Dietmar Göhlichb
Universität Berlin, Chair of Transport Systems Planning and Transport Telematics, Straße des 17. Juni 135, 10623 Berlin, Germany
abTechnische
TechnischeUniversität
UniversitätBerlin,
Berlin,Chair
Chairof
ofTransport Systems
Methods for Planning
Product and Transport
Development Telematics,Straße
and Mechatronics, Straßedes
des17.
17.Juni
Juni135,
135,10623
10623Berlin,
Berlin,Germany
Germany
b Technische Universität Berlin, Chair of Methods for Product Development and Mechatronics, Straße des 17. Juni 135, 10623 Berlin, Germany

Abstract
Abstract
Decarbonisation is a major challenge for the coming decades, for all industries, including the transport sector. Battery electric
Decarbonisation is a major
vehicles are a potential challenge
solution for thefortransport
the coming decades,
sector to reducefor its
all carbon
industries, including
impact. Asidesthe transport
from sector. whether
the question Battery there
electric
is
vehicles are a potential solution for the transport sector to reduce its carbon impact. Asides from the question
sufficient supply of electric vehicles for freight transport, it is also unclear whether battery-powered trucks meet the practical whether there is
sufficient supply
requirements, of electric
especially vehicles
in terms for freight
of their drivingtransport,
range. Toitinvestigate
is also unclear whether tours
this, synthetic battery-powered
were generatedtrucks bymeet the apractical
solving Vehicle
requirements,
Routing Problem especially
(VRP).inThisterms of generates
also their driving the range. To and
fleet size investigate this, synthetic
composition depending tours
on were
a set generated
of differentbyvehicle
solvingtypes.
a Vehicle
The
Routing Problem (VRP). This also generates the fleet size and composition depending on a set of different
network with underlying traffic conditions comes from an publicly available transport model. The generated tours are then simulated vehicle types. The
network
with with underlying
an open-source trafficsimulation
transport conditions(MATSim),
comes fromfor anboth
publicly available
diesel transport
and battery model.
electric The (BEVs).
vehicles generatedIntours are then study,
a sensitivity simulated
two
with an open-source
different transport
purchase prices weresimulation
considered(MATSim), for both
for calculating diesel
vehicle and The
costs. battery
caseelectric vehicles
study uses (BEVs).
a model In afood
of the sensitivity
retailingstudy, two
industry
different purchase prices were considered for calculating vehicle costs. The case study uses a model of the
for the city of Berlin. 56% of the tours can be driven without recharging. When recharged one time, 90% of the tours are suitable food retailing industry
for BEVs.
for the cityThe
of Berlin. 56%
costs for of the tours
transporting thecan
goodsbe driven withoutbyrecharging.
will increase 17 to 23% When recharged
depending on theone time, 90%
assumption forofthethepurchase
tours areprices
suitable
for
for
the BEVs.
BEVs. The
Usingcosts for transporting
a well-to-wheel the goodsthe
calculation, will increase by of
electrification 17all
to tours
23% depending on the assumption
leads to a reduction for the
of greenhouse gaspurchase prices for
(GHG) emissions
the BEVs.
by 26 Using
to 96% a well-to-wheel
depending calculation,
on the assumed the electrification
electricity production.of all tours leads to a reduction of greenhouse gas (GHG) emissions
by 26 to 96% depending on the assumed electricity production.
c 2020

© 2020 The Authors. Published
The Authors. Published byby Elsevier
Elsevier B.V.
B.V.
c 2020

This The Authors. Published by Elsevier B.V.
This is an open access article under the
is an open access article under the CC
CC BY-NC-ND
BY-NC-ND licenselicense(http://creativecommons.org/licenses/by-nc-nd/4.0/)
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
This is an open
Peer-review
Peer-review access
under article under
responsibility theConference
the
of the CC BY-NC-ND
Conference license
Program
Program (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Chairs.
Chairs.
Peer-review under responsibility of the Conference Program Chairs.
Keywords: freight transport; decarbonisation, electrification; agent-based modelling; MATSim; vehicle routing problem
Keywords: freight transport; decarbonisation, electrification; agent-based modelling; MATSim; vehicle routing problem

1. Introduction
1. Introduction
Reducing greenhouse gas emissions to limit global warming and climate change with all its consequences is one of
theReducing greenhouse
major global gasof
challenges emissions to limit global
the 21st century. warmingtoand
As a response thisclimate change
challenge, thewith all its Commission
European consequencesagreed
is one on
of
the major global challenges of the 21st century. As a response to this challenge, the European Commission
the ”European Green Deal”, which includes having net zero emissions of greenhouse gases (GHG) by 2050 and de- agreed on
the ”European
coupling of the Green Deal”,
economic which
growth includes
from having
resource net To
use [6]. zeroachieve
emissions of greenhouse
this goal, gases (GHG)
”a 90% reduction by 2050
in transport and de-
emissions
coupling of the economic growth from resource use [6]. To achieve this goal, ”a 90% reduction in transport emissions

∗ Corresponding author. Tel.: +49-30-314-29592 ; fax: +49-30-314-26269

∗ Corresponding
E-mail address:author. Tel.: +49-30-314-29592 ; fax: +49-30-314-26269
[email protected]
E-mail address: [email protected]
1877-0509  c 2020 The Authors. Published by Elsevier B.V.
1877-0509
This c 2020

is an open Thearticle
access Authors. Published
under by Elsevier B.V.
the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
1877-0509
This
Peer-review ©
is an open2020
under Thearticle
access Authors.
under
responsibility Published
of by Elsevier
the Conference
the CC BY-NC-ND B.V. Chairs.
license
Program (http://creativecommons.org/licenses/by-nc-nd/4.0/)
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review under responsibility of the Conference Program Chairs.
Peer-review under responsibility of the Conference Program Chairs.
10.1016/j.procs.2020.03.159
758 Kai Martins-Turner et al. / Procedia Computer Science 170 (2020) 757–763
2 K. Martins-Turner, A. Grahle, K. Nagel, D. Göhlich / Procedia Computer Science 00 (2019) 000–000

is needed by 2050” [6]. Currently 35% of CO2 emissions in Germany are emitted by trucks [2]. Besides reducing the
transport system’s impact on global climate, bans on diesel vehicles are being discussed in various cities to protect the
population from harmful emissions[27, 30]. In this paper, we investigate if and to what extent the current requirements
for urban freight transport can be fulfilled when replacing the current internal combustion engine vehicles (ICEVs) by
battery electric vehicles (BEVs).
Goods for a city typically arrive at distribution centers (“hubs”) at the periphery of the city. From these hubs, they
are distributed throughout the city. Generating tours for a fleet to perform this distribution task can be done by solving
the Vehicle Routing Problem (VRP). An open source toolkit for solving VRPs is jsprit [17], which can be used stand
alone as well as in conjunction with the agent-based transport planning software MATSim [32].

Vehicle Routing Problem (VRP). For transport companies, creating optimal vehicle routes depends on various factors.
These factors can be internal and external. The internal factors include the location of depot(s) and the available fleet
at the depot(s). Each vehicle type in a fleet has a set of given properties, e.g. carrying capacity and fixed and variable
costs [17]. The most important external specifications are customer requests: quantity and type of goods, location and
time window to deliver goods. Other external constraints are the road network and (expected) traffic situations, tolls
and driving restrictions [3]. Consideration of congestion is usually not part of urban tour planning [5].
A VRP can be defined as follows: finding a plan to ”determine a set of vehicle routes to perform [...] transportation
requests with the given vehicle fleet at minimum costs” [15]. Solving a VRP answers two main sub-problems for the
carrier: (i) assigning requests to tours (clustering) and (ii) determining the optimal sequence in which the requests are
served within a tour (routing) [21, 25]. More information regarding different types of vehicle routing problems can
found e.g. in [28], [25] or [17].

jsprit. For solving VRPs, several toolkits are available. One of these is the open-source toolkit jsprit [17]. Jsprit
optimizes the solution iteratively by ruining parts of the solution and recreating them. The objective function is to
reduce the costs for the carriers. Inputs are fixed costs for the vehicle including depreciation, insurance, maintenance
etc., variable cost for distance and travel time, and other costs such as tolls, or penalties for missed time windows [17].
Each VRP has a carrier with a certain fleet and list of requests. The fleet is located at the depot(s) and can be defined
either with concrete vehicle numbers, or with an infinite fleet of specified vehicle types. When using the infinite fleet
size, the fleet composition is also part of the solution. The requests (demand) can be defined using one location (the
other one is the depot) or explicitly by defining pickup and delivery location. Defining only one location opens the
possibility to solve a multi-depot VRP (MDVRP), while defining both pickup and delivery locations allows the vehicle
to reload goods during the tour. Further information about both kinds of defining the requests can be found in [18].

2. Methodology

This study uses the open source software Multi-Agent Transport Simulation (MATSim, https://matsim.org,
https://github.com/matsim-org/matsim). It is an activity-based, extendable framework that is designed for
agent-based transport simulations of large-scale scenarios. Several optional extensions are available (https://
matsim.org/extensions) as well as open-access scenarios (https://matsim.org/open-scenario-data) [12].
Investigating the effects of measures is a three-step approach: (1) Building a model of the base case, (2) building a
model of the policy case(s) and (3) comparing the results, e.g. costs and benefits.

Simulation model. The model of the base case is distributing goods to the food retailing shops with ICEVs, in this
case diesel powered trucks. For this, a model of the daily demand is needed as well as a method to generate plausible
vehicle tours serving that demand. With some modifications, the demand is based on earlier studies (see section 3).
The vehicle tour for each vehicle starts at a depot, going back and forth between depot and delivery locations (shops)
and finally returns to the depot. These tours need to be generated in advance of the traffic simulation, because the
agents (persons, drivers) in the traffic simulation have to follow their tour plan [32]. For generating the tour plan, we
are using the open source software jsprit [17] as heuristic Vehicle Routing Problem (VRP) solver. It is integrated into
MATSim using the MATSim freight contrib [32]. Because of this integration we are, in contrast to [5], able to consider
traffic congestion in our model (see section 3).
 Kai Martins-Turner et al. / Procedia Computer Science 170 (2020) 757–763 759
K. Martins-Turner, A. Grahle, K. Nagel, D. Göhlich / Procedia Computer Science 00 (2019) 000–000 3

For the policy cases, the same demand is assumed. However, now only battery electric vehicles (BEVs) are allowed
to transport the goods. Running the scenario means that we solve the VRP using jsprit with 10 000 iterations and
afterwards run a single MATSim iteration. From the output events, we can calculate the travel times and distances for
each vehicle. In an ex-post analysis, we analyse the effects on the carriers in terms of fleet compositions and costs.
We also evaluate the tours of the vehicles to see under which conditions it is possible to use battery electric trucks as
replacement for diesel trucks.

Well-to-Wheel (WTW) analysis. To analyze the environmental impact of the simulated scenarios, GHG emissions
from the production of diesel and electricity as well as from their use in the vehicles are estimated following
the well-to-wheel (WTW) methodology [16]. We calculate GHG emissions from electricity production assuming
518 gCO2eq/kWh for the electricity production in 2018 [13]. Furthermore, we are assuming 347 gCO2eq/kWh
for the electricity production in 2030 and 25 gCO2eq/kWh for renewable energies [29]. For the diesel we assume
3 170 gCO2eq/l diesel [4].

3. Case Study

This case study is located in Berlin, the capital and largest city of Germany. It deals with transporting goods to the
shops of the food retailing industry. It is based on the case study of Schröder and Liedtke [26].

Road Network. The network is taken from the MATSim Open Berlin scenario (https://github.com/
matsim-scenarios/matsim-berlin), which is available publicly and free of charge [31]. Since we are only inter-
ested in the effects of changing the truck fleet from ICEVs to BEVs, we do not run a simulation that includes passenger
transport. The time dependent travel times in the network were generated accordingly to the observed traffic in the
simulation output of the MATSim Open Berlin Scenario, using so called NetworkChangeEvents [24].

Freight demand. For this scenario we use the demand model of Schröder and Liedtke [26], together with modifica-
tions by [18]. The main modification is enabling vehicles to run more than one delivery tour from the depot. This
implies that loading vehicles with goods at the depot becomes part of the VRP. We assume the same average loading
time of three minutes per pallet at the depot as it is assumed by [26] for unloading one pallet. In total, the demand of 15
retailing companies in Berlin with 17 distribution centers located in and around Berlin is modelled. These companies
serve 1057 food retailing shops with 1928 demand requests. The shops’ demand is the demand of an average day.
Each shop has a variety of heterogeneous products they offer to their consumers. Schröder and Liedtke aggregated
the different products a shop offers to their consumers into three different groups: fresh, frozen, and dry. These groups
require different transport conditions and thus were transported by different carriers in the model [26]. This leads to
groups
15 companies · 3 company = 45 carriers in the model.
Since the vehicles can run several (sub-)tours that include reloading goods during a day, the original time-window
constraint of the model by [26] was modified. The vehicles availability as well as the delivering time-windows for dry
and frozen goods is set to [9am – 7pm]. The time window for fresh goods remains unmodified [4am – 9am]. [18]
Each carrier plans its tours based on its own fleet available at the depots. The available vehicle types in the fleet of
a carrier depend on the type of goods.

Vehicle types and costs calculation. In the scenario of Schröder and Liedtke, some assumptions about the available
vehicle types were made. They define four different vehicle types with the size of the vehicles as distinguishing
criterion: light 7.5 tons, medium 18 tons, heavy 26 tons and heavy 40 tons. All vehicles are ICEVs. All vehicles are
located at a depot, but not all vehicle types are available for each carrier [26, 9]. The number of vehicles of the different
available types are not limited for each carrier. Therefore, the fleet composition is a result of the VRP.
The cost calculation from the study [26] has been updated. The used methodology and many values for the cal-
culation are now based on the German Bundesverkehrswegeplan (BVWP, Federal Transport Plan [1]). The purchase
cost of the vehicles is depreciated half by time and half by driven distance. Data for representative vehicle types are
given in [20]; the numbers include driver wages. Because the BVWP models from a national economics perspective,
taxes and insurance premiums are not included in the cost calculation of BVWP [22]. However, insurances premiums
and other taxes except the sales tax (VAT) are relevant for the carrier and should therefore be included in the costs for
760 Kai Martins-Turner et al. / Procedia Computer Science 170 (2020) 757–763
4 K. Martins-Turner, A. Grahle, K. Nagel, D. Göhlich / Procedia Computer Science 00 (2019) 000–000

tour planning. Therefore, missing values for insurances and taxes were taken from [7]. We use the following vehicle
types in the base scenario (ICEVs):

• light – 7.5 tons: Mercedes Atego 818L / MAN TGL 12.220 BL

• medium – 18 tons: MAN TGX 18.440 XLX
• heavy – 26 tons: Mercedes Actros 2544 LL
• heavy with trailer – 40 tons: MAN TGM 18.340 BL + Trailer 7.75m

Battery Electric Trucks (BEV). For this study, the battery size of the electric trucks is selected in a way that the
maximum payload is equal for the ICEVs and BEVs. It has to be taken into account that according to [19] within the
European Union an additional ton gross vehicle weight (GVW) is allowed for medium and heavy trucks and up to two
additional tons for heavy trucks with a trailer when using a clean propulsion system. We assume that only 70% (i.e.
from 10% to 80%) of the theoretical battery capacity is used in order to ensure an adequate battery life. According
to current publications on battery management such as [23], this is rather conservative. To calculate the possible
ranges, consumption data for electric trucks currently available on the market are chosen. This leads to consumption
specifications of 61 kWh/100km for the light, 106 kWh/100km for the medium, 150 kWh/100km for the heavy and
180 kWh/100km for the heavy truck with trailer. These specifications are supported by findings of current publications
such as [10]. With these specifications, we calculate possible distance ranges of 100 km for the 7.5 tons truck, 80 km
(18 tons), 133 km (26 tons) and 172 km (40 tons).
The costs for electric vehicles are strongly driven by the battery. Therefore, we consider the prices for the chassis
and for the battery separately. For the chassis, market analysis shows that current electric chassis are about 1.6 times
more expensive than ICEV chassis. According to [8], the prices for a BEV chassis will approach the prices for ICEV as
soon as mass production picks up pace. Therefore we will consider the current situation with a 60% price increase from
ICEV to BEV chassis (case A: BEV 160) and the possible future situation with equal chassis prices as a sensitivity
analysis (case B: BEV 100).
The cost for commercial vehicle batteries is significantly higher than batteries for private vehicles [14]. Following
[14] and our own analysis of the current market for commercial BEVs we choose a battery price of 600 e/kWh.
Furthermore, we set the possible lifetime charging cycles to 4000 and specific energy density to 0.15 kWh/kg on pack
level according to [11]. Due to the possible charging cycles, no battery change is required in the presumed use phase
of the vehicles, even with an average of more than one charging event per day.
The resulting costs structure for all vehicle types used as input for the VRP is summarized in table 1.

Table 1. Cost parameters for the different vehicle types

vehicle type cost type Base: ICEV A: BEV 160 B: BEV 100

fixed [e/day] 63.49 81.04 74.76

7.5 tons variable per distance [e/m] 0.00040 0.00051 0.00046
variable per time [e/s] 0.00490 0.00490 0.00490
fixed [e/day] 80.47 107.43 96.26
18 tons variable per distance [e/m] 0.00065 0.00061 0.00055
variable per time [e/s] 0.00490 0.00490 0.00490
fixed [e/day] 82.60 107.43 111.90
26 tons variable per distance [e/m] 0.00067 0.00076 0.00072
variable per time [e/s] 0.00490 0.00490 0.00490
fixed [e/day] 126.58 192.80 183.93
40 tons variable per distance [e/m] 0.00069 0.00080 0.00078
variable per time [e/s] 0.00559 0.00559 0.00559

4. Results

As explained in Sec. 2, the fleet and tour planning is run for the retail scenario and the three different vehicle
configurations: ICEV, BEVs 160, and BEVs 100. It is possible that vehicles can reload goods at the depot and run one
or more additional sub-tour(s). The resulting daily tours are analysed ex-post after the traffic simulation.
 Kai Martins-Turner et al. / Procedia Computer Science 170 (2020) 757–763 761
K. Martins-Turner, A. Grahle, K. Nagel, D. Göhlich / Procedia Computer Science 00 (2019) 000–000 5

Table 2. Results of the simulation for a typical workday

Base: ICEV A: BEV 160 B: BEV 100

Costs [e] 67 288 83 087 78 659

total 283 275 279
light 7.5 tons 17 20 21
vehicles used medium 18 tons 16 24 27
heavy 26 tons 223 207 207
heavy 40 tons 27 24 24
Distance travelled [km] 36 783 37 368 37 761
Time travelled [h] 1 054 1 074 1 070

Figure 1 shows the distance of the tours driven in the network for the three cases. The distances in both policy
cases seems with few exceptions comparable to the distances observed in the base case. The BEV tours are planned
without range constraint (see Sec. 2). It is assumed that any necessary recharging could be done during pickup or
drop-off activities without additional costs. However, 56% of all tours could be driven without recharging during the
day. Assuming one recharging during the tour and therefore doubling the vehicles’ range would result in 90% possible
tours with BEVs (see Table 3).
Table 2 shows some results of the simulation for a typical workday. In terms of total costs for the carriers, we
observe an increase by or approx. 15 800 e/day (23.5%) when using BEVs compared to ICEVs in the case that the
chassis prices of the BEV will be by 60% higher as they are for ICEVs (A). In the sensitivity case (B) with equal
chassis prices, the cost for the carrier will increase by 11 713 e/day (16.9%). The reason why the cost increases are
not larger is that personnel costs are a large portion of overall costs, and they do not change between the cases.
Distance and time travelled increases by 1.5 to 2.7%. This is presumably related to a small reduction of the fleet
size. The decreasing fleet size is reasonable since higher costs for the BEVs compared to the ICEVs forces jsprit to a
more efficient solution regarding the vehicle usage.
The WTW GHG emissions of the three different simulation cases are calculated by multiplying the observed
distance travelled (see Tab. 2, Fig. 1) with the vehicle type specific energy consumption and the specific CO2 emissions
factor per energy unit. The diesel consumption for the ICEVs were extracted from [20] and goes from 13.57 l/100
km for the 8 tons truck to 37.45 l/100km for the 40 tons truck. The vehicle type specific energy consumption for
the BEVs is between 61 and 180 kWh/100km as mentioned in Sec 2. Three different emission factors were used to
show the effects of electrification depending from the electricity production. For calculating the per year emissions,
250 workdays/year are assumed [20]. We can observe a reduction of the GHG emissions from approx. 9 500 t/year
using ICEVs to approx. 7 000 t/year (-26%) using BEVS and assuming electricity production in 2018. Assuming the
expected future German electricity production, approx. 4 700 t GHG/year (-50%) would be emitted with the BEVs.
With renewable energy, the GHG emission would decrease to approx. 340 t/year (-96%). As a consequence of the
very similar travel distances for both BEV cases the GHG emissions of both these cases are very similar as well.

5. Conclusion and Outlook

In this paper, we analysed whether electrification is suitable for the urban last mile supply of shops based on a case
study in the city of Berlin. The BEVs were designed to have the same payload as ICEVs in the same weight class. In

Table 3. Proportion of tours that can be driven with BEVs according to i) single and ii) doubled range
i: without recharging ii: with one-time recharging: range doubled
vehicle Type range A: BEV 160 B: BEV 100 range A: BEV 160 B: BEV 100

7.5 tons 100 km 9 / 20 45 % 9 / 21 43 % 200 km 20 / 20 100 % 20 / 21 95 %

18 tons 80 km 7 / 24 29 % 5 / 27 19 % 160 km 18 / 24 75 % 15 / 27 56 %
26 tons 133 km 131 / 207 63 % 137 / 207 66 % 266 km 202 / 207 98 % 203 / 207 98 %
40 tons 172 km 7 / 24 29 % 7 / 24 29 % 374 km 11 / 24 46 % 13 / 24 54 %
total 154 / 275 56 % 158 / 279 57 % 251 / 275 91 % 251 / 279 90 %
762 Kai Martins-Turner et al. / Procedia Computer Science 170 (2020) 757–763
6 K. Martins-Turner, A. Grahle, K. Nagel, D. Göhlich / Procedia Computer Science 00 (2019) 000–000

Fig. 1. Observed daily tour distances [km] driven by vehicle type and case. Fig. 2. Calculated CO2 emissions per year.

the selected scenarios we could show with an ex-post analysis that 56% percent of the resulting tours can be driven
with battery electric trucks without recharging during the tour, another 34% with recharging one time.
In terms of costs, the change from ICEVs to BEVs will result in increasing costs by 23% if the chassis of the BEVs
are 60% more expensive as the chassis of the ICEVs (case A). In the sensitivity case (B) with equal chassis cost, the
total transportation costs will increase by 17%. These costs include the costs for the battery as well as the costs for
diesel respectively electric energy. Regarding the fleet size and composition, no significant change is observed.
Using BEVs for all tours leads to a reduction of 2 500 to 9 200 t GHG per year depending on the assumed electricity
production. This is a reduction by 26 to 96% compared to the usage of ICEVs.
In this paper we evaluate the feasibility of using BEVs with an ex-post analysis. Including a vehicle type specific
distance constraint into the VRP can force respecting this limitation already in the tour generation. Based on this,
including (re-)charging into the VRP solving may be a useful further development. It may also force to replan the
logistics network, introducing additional hubs between the depots and the shops. Providing additional variants of
vehicles with larger battery capacity is another option, even though this leads to a reduction of the payload. Regardless
of these improvement potentials for our method, we were able to show that even without major changes in the carriers
operating strategies, a significant amount of urban freight delivery tours can be electrified with current technology.

Acknowledgements

This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – project
number: 398051144. We thank Stefan Schröder and Gernot Liedtke for providing their data of the Berlin food retailing
scenario.

References

[1] BMVI, 2016. Bundesverkehrswegeplan 2030. http://www.bmvi.de/DE/Themen/Mobilitaet/

Infrastrukturplanung-Investitionen/Bundesverkehrswegeplan-2030/bundesverkehrswegeplan-2030.html.
[2] Bundesministerium für Umwelt, Naturschutz und nukleare Sicherheit (BMU), 2018. Klimaschutz in Zahlen. Fakten, Trends und Impulse
deutscher Klimapolitik. Ausgabe 2018.
[3] Cerwenka, P., 2007. Handbuch der Verkehrssystemplanung. Österreichischer Kunst- und Kulturverlag.
[4] DIN EN 16258:2012, 2013. Methodology for calculation and declaration of energy consumption and GHG emissions of transport services
(freight and passengers) (DIN EN 16258:2012); German version EN 16258:2012 [Methode zur Berechnung und Deklaration des Energie-
verbrauchs und der Treibhausgasemissionen bei Transportdienstleistungen (Güter- und Personenverkehr) (DIN EN 16258:2012); Deutsche
Fassung EN 16258:2012]. Deutsches Institut für Normung e.V. 03/2013.
[5] Ehmke, J.F., 2012. Integration of information and optimization models for routing in city logistics. volume 177 of International Series in
Operations Research & Management Science. Springer Science+Business Media.
[6] Europäische Kommission, 2019. Der europäische Grüne Deal. COM(2019) 640 final. COM(2019) 640 final.
[7] Eurotransport, 2017. Lastauto Omnibus Katalog 2018.
 Kai Martins-Turner et al. / Procedia Computer Science 170 (2020) 757–763 763
K. Martins-Turner, A. Grahle, K. Nagel, D. Göhlich / Procedia Computer Science 00 (2019) 000–000 7

[8] Fuessel, A., 2017. Technische Potenzialanalyse der Elektromobilität: Stand der Technik, Forschungsausblick und Projektion auf das Jahr 2025.
Research, Springer Vieweg, Wiesbaden. URL: http://www.springer.com/.
[9] Gabler, M., Schröder, S., Friedrich, H., Liedtke, G., 2013. Generierung der Nachfragestrukturen für die mikroskopische Simulation des
städtischen Distributionsverkehrs im Lebensmittelhandel, in: Clausen, U., Thaller, C. (Eds.), Wirtschaftsverkehr 2013. Springer Berlin Heidel-
berg, pp. 32–48. URL: http://dx.doi.org/10.1007/978-3-642-37601-6_3, doi:10.1007/978-3-642-37601-6_3.
[10] Gao, Z., Lin, Z., Franzese, O., 2017. Energy consumption and cost savings of truck electrification for heavy-duty vehicle applications.
Transportation Research Record: Journal of the Transportation Research Board 2628, 99–109. doi:10.3141/2628-11.
[11] Göhlich, D., Fay, T.A., Park, S., 2019. Conceptual design of urban e-bus systems with special focus on battery technology. Proceedings of the
Design Society: International Conference on Engineering Design 1, 2823–2832. doi:10.1017/dsi.2019.289.
[12] Horni, A., Nagel, K., Axhausen, K.W. (Eds.), 2016. The Multi-Agent Transport Simulation MATSim. Ubiquity, London. doi:10.5334/baw.
[13] Icha, P., Kuhs, G., 2019. Entwicklung der spezifischen Kohlendioxid-Emissionen des deutschen Strommix in den Jahren 1990 – 2018. Tech-
nical Report Climate Change 10/2019. Umweltbundesamt.
[14] IKT für Elektromobilität, 2015. Gesamtbericht Wirtschaftlichkeit von Elektromobilität.
[15] Irnich, S., Toth, P., Vigo, D., 2014. Vehicle routing: problems, methods, and applications. chapter 1 The Family of Vehicle Routing Problems.
in: [28]. pp. 1–33. doi:10.1137/1.9781611973594.ch1.
[16] JRC, R.E., HASS, H., LARIVÉ, J.F., JRC, L.L., MAAS, H., Rickeard, D., 2014. Well-to-wheels report version 4. a jec well-to-wheels analysis.
Institute for Energy and Transport, Joint Research Centre, Luxembourg: Publications Office of the European Union 2014.
[17] jsprit, accessed 02-dez-2018. https://github.com/graphhopper/jsprit.
[18] Martins-Turner, K., Nagel, K., 2019. How driving multiple tours affects the results of last mile delivery vehicle routing problems. Procedia
Computer Science 151, 840–845.
[19] PARLIAMENT, T.E., UNION, T.C.O.T.E., 29.04.2015. Directive (eu) 2015/719 of the european parliament and of the council. URL: http:
//data.europa.eu/eli/dir/2015/719/oj.
[20] Planco, ITP, TUBS, 2015. Grundsätzliche Überprüfung und Weiterentwicklung der Nutzen-Kosten-Analyse im Bewertungsverfahren der
Bundesverkehrswegeplanung. Endbericht FE Projekt Nr. 960974/2011. Planco GmbH, Intraplan Consult GmbH, TU Berlin Service GmbH.
Im Auftrag des BMVI. Auch VSP WP 14-12, see http://www.vsp.tu-berlin.de/publications.
[21] Prockl, G., 2010. Informationsmanagement, in: Stölzle, W. (Ed.), Güterverkehr kompakt. Oldenbourg. Lehrbuch kompakt. chapter Informa-
tionsmanagement, pp. 151–165.
[22] PTV, TCI, Mann, H.U., 2016. Methodenhandbuch zum Bundesverkehrswegeplan 2030. http://www.bmvi.
de/DE/VerkehrUndMobilitaet/Verkehrspolitik/Verkehrsinfrastruktur/Bundesverkehrswegeplan2030/
InhalteHerunterladen/inhalte_node.html.
[23] Rehman, M.M.U., Zhang, F., Evzelman, M., Zane, R., Smith, K., Maksimovic, D., 18.09.2016 - 22.09.2016. Advanced cell-level control
for extending electric vehicle battery pack lifetime, in: 2016 IEEE Energy Conversion Congress and Exposition (ECCE), IEEE. pp. 1–8.
doi:10.1109/ECCE.2016.7854827.
[24] Rieser, M., Nagel, K., Horni, A., 2016. Matsim data containers, in: [12]. chapter 6. pp. 55–60. doi:10.5334/baw.
[25] Scheuerer, S., 2004. Neue Tabusuche-Heuristiken für die logistische Tourenplanung bei restringierendem Anhängereinsatz, mehreren Depots
und Planungsperioden. phdthesis. Universität Regensburg. URL: https://epub.uni-regensburg.de/10196/.
[26] Schröder, S., Liedtke, G., 2014. Modeling and analyzing the effects of differentiated urban freight measures – a case study of the food retailing
industry. Annual Meeting Preprint 14-5015. Transportation Research Board. Washington D.C.
[27] Senatsverwaltung für Stadtentwicklung und Umwelt (SenStadt), 2013. Luftreinhalteplan 2011 bis 2017 für Berlin.
[28] Toth, P., Vigo, D. (Eds.), 2014. Vehicle routing: problems, methods, and applications. Society for Industrial and Applied Mathematics (SIAM).
doi:10.1137/1.9781611973594.
[29] Wietschel, M., Kühnbach, M., Rüdiger, D., 2019. Die aktuelle Treibhausgasemissionsbilanz von Elektrofahrzeugen in Deutschland. Technical
Report. Working Paper Sustainability and Innovation.
[30] Wikipedia, accessed 09-jan-2019. Dieselfahrverbot. https://de.wikipedia.org/wiki/Dieselfahrverbot.
[31] Ziemke, D., Kaddoura, I., Nagel, K., 2019. The MATSim Open Berlin Scenario: A multimodal agent-based transport simulation scenario
based on synthetic demand modeling and open data. Procedia Computer Science 151, 870–877. doi:10.1016/j.procs.2019.04.120.
[32] Zilske, M., Joubert, J.W., 2016. Freight traffic, in: [12]. chapter 24. pp. 155–156. doi:10.5334/baw.