Adapting Retail Supply Chains for the Race to Sustainable Urban

Delivery

Angie Ramirez-Villamil

1,2 a

, Anicia Jaegler

2 b

and Jairo R. Montoya-Torres

1 c

School of Engineering, Universidad de La Sabana, Km 7 Autopista Norte de Bogot

a, D.C., Ch

ıa, Colombia

Kedge Business School, 40 Avenue des Terroirs de France, 75012 Paris, France

Keywords:

Urban Logistics, Supply Chain Management, Sustainability, Two-Echelon Distribution, Cargo Bikes, Retail.

Abstract:

To deal with urban distribution challenges, companies are redesigning their distribution networks. This paper

studies a two-echelon vehicle routing problem, one of the most employed models, with a heterogeneous ﬂeet

between echelons. Vehicles in the ﬁrst echelon are mobile satellites that supply the vehicles in the second

echelon. Our study aims to minimize the travel time. To solve this complex problem when facing real-life

distribution, a heuristic solution approach is followed by decomposing the components of the problem and

applying the well-known nearest neighbor procedure. This approach is also justiﬁed by the very large amount

of delivery points, so the problem dataset can be computationally tractable. Experiments are run using real

data from a delivery company in Paris, France. Different scenarios are evaluated, and results show that the

consideration of cargo bikes has big potential to reduce some of the externalities caused by conventional

delivery systems, while some non-intuitive impacts are also found, such as the increase in land use.

1 INTRODUCTION

Internet shopping offers more shopping possibilities

to consumers, and an additional distribution chan-

nel to retailers (Kull et al., 2007). The pandemic

of COVID-19 has reinforced this trend by demand-

ing supply chain managers to rethink their operations

(Montoya-Torres et al., 2021). Supply chain manage-

ment are the core of activities that are relevant to e-

commerce to support its exponential growth. How-

ever, its complexity has considerably boost since sev-

eral operations (e.g., warehousing, inventory, pack-

aging, product shipping and tracking) have begun to

be considered (Sandhaus, 2019). In addition, de-

mand uncertainty has increased due to its accelerated

growth. Retailers and suppliers should start looking

for different initiatives to minimize stock-outs and

guarantee service levels (Bendoly et al., 2018).

On another side, urbanization is a constant trend

generating problems in freight transportation due to

the delivery of online retail orders. Their increase,

the constraints linked to urban environments and poli-

cies, and environmental pressures force practitioners

https://orcid.org/0000-0001-6840-3525

https://orcid.org/0000-0002-3014-4561

https://orcid.org/0000-0002-6251-3667

to rethink traditional deliveries. Indeed, urban freight

transportation has become an important component

of urban planning. For instance, carriers have been

challenged to provide higher levels of service at lower

costs to satisfy customers’ needs, such as same-day-

delivery (Stroh et al., 2022). They have made efforts

to organize their freight transport systems as they ob-

struct themself and other road users by causing con-

gestion during loading/unloading operations, with as-

sociated negative environmental impacts (air pollu-

tion and noise). In addition, current environmental

agreements, as well as low-emission urban areas, are

requiring freight carriers to reduce the CO

equivalent

(CO

e) emissions.

City logistics initiatives and strategies have been

developed and modeled to improve efﬁciency, re-

lieve trafﬁc congestion, and reduce CO

e emissions.

One is the redesign of urban distribution networks by

adding intermediate nodes i.e., hubs, satellites, urban

logistic spaces, located in the proximity of an urban

area and allowing the consolidation of freight ﬂows

(Meza-Peralta et al., 2020; Browne et al., 2005). To

make last-mile delivery more efﬁcient. Hence, signif-

icant efforts have been made to design efﬁcient op-

timization models and algorithms capable of support-

ing logistics decision makers (Ramirez-Villamil et al.,

2022; Patier and Browne, 2010).

354

Ramirez-Villamil, A., Jaegler, A. and Montoya-Torres, J.

Adapting Retail Supply Chains for the Race to Sustainable Urban Delivery.

DOI: 10.5220/0012797800003758

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 14th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2024), pages 354-362

ISBN: 978-989-758-708-5; ISSN: 2184-2841

When designing the last-mile supply network, the

most employed model is the two-echelon distribu-

tion system, which consists of delivering goods from

one depot to a set of satellites and from there to a

set of customers geographically dispersed within an

urban area (Marrekchi et al., 2021). When dealing

with routing decision, this problem is modeled as a

two-echelon capacitated vehicle routing problem (2E-

CVRP) (Gonzalez-Feliu et al., 2008). It is known to

be NP-hard, which means that the time needed to ﬁnd

an optimal solution grows exponentially in function

of the size of the instance. Thus, many researchers are

motivated to seek approximate algorithms (heuristics

and metaheuristics) to solve it. The importance and

attractiveness of this research topic concerns its ap-

plicability in real cases of freight transportation plan-

ning, the research opportunity in terms of the devel-

opment of efﬁcient solution methods, the use of novel

disruptive delivery technologies, and the inclusion

of more realistic features (e.g., satellite synchroniza-

tion, use of environmentally friendly vehicles, multi-

ple depots, mobile depots, and considering very large

amount of delivery points) (Carlsson and Song, 2018;

Reed et al., 2022).

In the context of more sustainable deliveries, de-

livery by cargo bikes is booming, as a preferred so-

lution for environmentally oriented decision-makers

(Silva et al., 2023). However, they are constrained in

terms of distance. This paper focuses on the imple-

mentation of mobile urban storage via delivery vans

loading the goods in the satellites for further last-mile

delivery by cargo bikes. Delivery time was also con-

sidered as part of the total travel time, to make the

problem more realistic. To the best of our knowl-

edge, there is no literature that addresses not only a

two-echelon delivery network using mobile satellites

to transport parcels until certain points within the city

where cargo bikes are loaded to perform last-mile de-

livery; but also, that includes the delivery time as part

of the total travel time to make the problem more re-

alistic. The problem includes both the transport from

the depot to satellites (ﬁrst echelon) and from these

mobile satellites to customers (second echelon). This

delivery network can contribute to the reduction of the

travel distance, and of the number of trucks in cities,

and consequently, it could decrease congestion.

This paper proposes a decomposition heuristic

based on the Nearest Neighbor procedure (NN). As

a real-life case study of a company delivering in the

city of Paris, France is considered, the rationale of

choosing a heuristic to solve the routing problem is

justiﬁed by the very large amount of delivery points.

Indeed, although research on VRP has witnessed a va-

riety of solution approaches, including exact, heuris-

tic and meta-heuristic algorithms (e.g., (Cattaruzza

et al., 2017; Vega-Mej

ıa et al., 2019; Soeffker et al.,

2022)), the literature has also highlighted the need of

fast yet comprehensive solution algorithms for easy

understanding in practice, instead of black-box com-

plex algorithms (Juan et al., 2015). As an objective

function, the minimization of travel times is consid-

ered. Key performance indicators like, CO

e and ﬁne

particles emissions, as well as ﬁxed cost and land use

are also addressed.

The approach is tested over different scenarios

simulating the actual options decision-makers may

encounter in real life. Hence, two scenarios for urban

freight distribution networks are proposed.

The remainder of this paper is organized as fol-

lows. Existent literature is ﬁrstly reviewed, followed

by description of the case study. The solution ap-

proach is then described, as well as the results ob-

tained. The paper ends with some conclusions, ex-

plaining some managerial implications, and suggest-

ing future research lines.

2 RELATED LITERATURE

The Internet has created opportunities for retailers to

increase sales (Nguyen et al., 2019), causing an evolu-

tion with the constant expansion of e-commerce. This

phenomenon has generated growth in online retail by

around an average of 10% each year from many years

and approximately 14% in western Europe in 2020

(Lone et al., 2021). Currently, online retailers offer a

variety of delivery options that are the result of com-

bining features such as: delivery speed, time slot, day

or overnight delivery, delivery date and delivery fee.

Over the years, there have been efforts to improve

and make online retail operations more efﬁcient with

the aim of generating value for the customers. (Buijs

et al., 2016) illustrated the importance of leveraging

opportunities in the design and control of a retail dis-

tribution network. The results obtained with the sim-

ulation model applied in a case study in The Nether-

lands show a reduction in travel distance by more than

40% when applying their multi-echelon approach and

demonstrate how even small changes in the distribu-

tion network design can lead to signiﬁcant improve-

ments in cross-docking performance.

As pointed out before, one of the most widely used

models in urban freight delivery is the 2E-VRP. The

problem has been studied since 1980 (Jacobsen and

Madsen, 1980). However, the ﬁrst study was pre-

sented by (Crainic et al., 2010), who promoted the use

of satellite platforms to redistribute goods in zones

where big trucks could not circulate due to the physi-

Adapting Retail Supply Chains for the Race to Sustainable Urban Delivery

355

cal or regulatory constraints, and showed a reduction

in the use of large vehicles by up to 72%. Since then,

this VRP variant has been extensively studied. In e-

commerce, (Zhou et al., 2018) proposed a multi-depot

two-echelon vehicle routing problem with delivery

options (MD-TEVRP-DO), where customers are al-

lowed to pick-up orders at intermediate facilities. Af-

ter applying a multi-population genetic algorithm, re-

sults showed that the ﬁnal cost can be reduced by 16%

(in comparison with the “no pick-up option”).

Nowadays, the use of different types of vehicles

in both ﬁrst and second echelons has become more

relevant in city logistics, but system becomes more

complex. In the ﬁrst echelon, trucks supply the satel-

lites, while in the second echelon, other transporta-

tion modes (e.g., cargo bikes, electric vehicles, UAVs,

vans, autonomous vehicles) perform the last-mile de-

liveries (Cattaruzza et al., 2017; Stamadianos et al.,

2023). Moreover, new variants of 2E-VRP have been

studied in sustainable applications. For example, dif-

ferent scenarios using cargo bikes for freight trans-

portation in inner-city areas have been proposed. (An-

derluh et al., 2019) present some of the successful

cases of cargo bikes in Europe for the cities Budapest,

Vienna, and Copenhagen. Other works also evaluated

the use of cargo bikes on different performance indi-

cators including cost and environmental impacts, to

analyze their use alone of as part of a mixed-ﬂeets

with electric vehicles i.e. (Caggiani et al., 2021).

The impact of cargo bikes in system performance was

shown to be very positive. However, none of previ-

ous work in the literature evaluate an extended set of

performance indicators, as in the current paper.

3 PROBLEM DESCRIPTION

A major French delivery company for the city of Paris

is taken as case study. It is a key player of the last-mile

and a leading brand in delivery, with the distribution

of approximately 63 million parcels. 95% of deliver-

ies are made on the ﬁrst attempt and 65% of parcels

are delivered directly to mailboxes. Its services are

used by major B2C clients, as well as B2B. For this

study, the company provided the data of 90,627 de-

liveries in Paris from four depots around Paris. Fig-

ure 1 presents the location of the four depots (pur-

ple points) and the location of the 90,627 customers

(green points). Furthermore, Paris is administratively

divided into 20 districts. The data provided are split

according to these districts. Actual demand and lo-

cation of points are kept conﬁdential. Such distribu-

tion problem is modeled as a two-echelon capacitated

vehicle routing problem (2E-CVRP) (Gonzalez-Feliu

et al., 2008). This problem, in its deterministic ver-

sion, is known to be NP-hard, which means that the

time needed to ﬁnd an optimal solution grows expo-

nentially in function of the number of delivery points.

The sub-problems that will be presented in the next

section depend on the scenarios presented in table 1.

Figure 1: Delivery points in Paris and four warehouses.

4 SOLUTION APPROACH

Since the 2E-CVRP is known for its hard complex-

ity (Gonzalez-Feliu et al., 2008), approximation al-

gorithms are good approaches to obtain feasible so-

lutions for medium- to large-sized instances in rea-

sonable computational time. Since the case study is a

very large-sized instance, a decomposition algorithm

(Figure 2) is proposed and the routing is solved using

the NN procedure (Taiwo et al., 2013). This algorithm

splits the problem into four sub-problems to reduce its

complexity, and aggregates the corresponding results

to guarantee the quality and feasibility of the solutions

and render it computationally tractable.

The ﬁrst subproblem is the random selection of

the location point for the satellites in each district of

Paris; the second one is to randomly cluster the satel-

lites to the depots; the third sub-problem is to ﬁnd a

set of routes starting from the depot to serve the cor-

responding satellites (ﬁrst echelon), and the last step

is to determine the routing from satellites to serve the

clients (second echelon).

4.1 Routing from Depots to Satellites

This subsection explains in a detailed way the heuris-

tic algorithm used to solve the ﬁrst three sub-

problems. The location of the satellites (ﬁrst sub-

problem) depend on the scenarios presented in table 1.

So, it is important to note that the number of satellites

and how they operate will depend on the scenario.

The second sub-problem consists of a random al-

location of the previously selected satellites to the

SIMULTECH 2024 - 14th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

356

Table 1: Differences between scenarios proposed.

Scenario 1 Scenario 2

From depots to satellites and from satellites to customers (last-mile delivery) From depots to nodes where mobile satellites meet cargo bikes for last-mile delivery

Homogeneous ﬂeet between echelons Heterogeneous ﬂeet between echelons

1st and 2nd echelon transportation mode: Delivery van (800 kg)

1st echelon: Delivery vans as mobile satellites (900 kg)

2nd echelon: Cargo bikes (180 kg)

One node randomly selected to be the satellite in each district of Paris The number of mobile satellites in the district depends on its demand and vehicle capacity

A total 20 satellite facilities with unlimited capacity A total of 96 mobile satellites with a limited capacity of 900 kg

Delivery time was considered

sc2-min: considers a delivery time of 2 min per client

sc2-max: considers a delivery time of 8 min per client

Figure 2: Flowchart of the proposed solution approach.

four depots (Figure 1). For scenario 1, it was decided

that 5 satellites would be assigned to each one of the

depots. While, for scenario 2, 5 districts would be

assigned to each one of the depots.

The third sub-problem is the ﬁrst-echelon routing

by an algorithm based on the NN procedure and fol-

lows the next steps:

1. Deﬁne the starting point of the route (Depot).

2. Find the nearest node to the last node added to the

path. If the nearest node is already in the path,

then choose the next closest.

3. Repeat step 2 until the vehicle reaches its maxi-

mum capacity.

4. Connect the last visited node to the depot to form

the tour. Calculate the distance traveled by the

vehicle and the total route time.

5. If there are unvisited nodes, add one more vehicle

and return to step 2.

4.2 Routing from Satellites to Clients

The last sub-problem is the design of routes for last-

mile delivery (second-echelon). For both scenarios,

satellites can only serve the nodes that belong to the

district assigned to them. It is important to note that

in scenario 1 the solution approach works in the same

way as the algorithm based on the NN procedure pre-

sented before. However, for scenario 2, as it has dif-

Adapting Retail Supply Chains for the Race to Sustainable Urban Delivery

357

ferent number of satellites for each district, the solu-

tion algorithm has some modiﬁcations. So, the prob-

lem is solved by following the next steps:

1. Divide the total number of clients of the district

into the number of satellites located into this dis-

trict.

2. Deﬁne the starting point of the route (Satellite i).

3. Find the nearest node to the last node added to the

path. If the nearest node is already in the path,

then choose the next closest.

4. Repeat step 3 until the vehicle n reaches its maxi-

mum capacity.

5. Connect the last visited node to the satellite to

form the tour. Calculate the distance traveled by

the vehicle and the total route time.

6. If there are unvisited nodes, add one more vehicle

and return to step 3.

7. If the total number of nodes for the satellite i were

already visited, the procedure for the satellite i + 1

is started (step 2).

8. Execute this procedure until all nodes are visited.

5 SCENARIOS SIMULATION

AND ANALYSIS OF RESULTS

The proposed solution procedure was coded in

Python, and experiments were run on a computer with

processor Intel® Core™ i7-10510U, CPU at 2.3 GHz

and 16 GB RAM. All below values were given by the

delivery ﬁrm. Two main scenarios are simulated. Ta-

ble 2 shows the distribution of the 96 satellites into the

20 districts of Paris for scenario 2. In this scenario,

delivery time was considered, which is the time that

the delivery man takes to unload the parcel, deliver

it, and return to the cargo bike to continue his route.

Different values were considered for the delivery time

from a minimum value of 2 minutes per delivery point

to a maximum of 8 minutes. Moreover, the working

hours of the delivery man are included, so the number

of trips required to visit all the delivery points with

5, 5.5 and 6 working hours is calculated. The value

presented for the number of tours is the average of the

three options of working hours.

Key performance indicators (Tables 3 and 4) such

as the number of vehicles, CO

e emissions, ﬁne parti-

cles, ﬁxed cost, and land use were considered to make

comparisons between scenarios. Mainly to assess if

the distribution network does not invade public space

in large proportions, if it is sustainable and if the ﬁxed

costs associated with the operation are low.

Table 2: Number of satellites per district in scenario 2.

District

N° of

satellites

District

N° of

satellites

District

N° of

satellites

District

N° of

satellites

D1 2 D6 3 D11 6 D16 8

D2 3 D7 4 D12 6 D17 7

D3 2 D8 5 D13 8 D18 5

D4 2 D9 4 D14 5 D19 3

D5 3 D10 4 D15 10 D20 6

Table 3: Fixed cost in euros per transportation mode.

Indicator / Parameter Cargo bike Vehicle

Private cost per km 0.13 0.45

Time value per hour 8.47 13.10

Revenue index 0.83% 1.00%

Time value per hour according to the revenue 7.03 13.09

Speed (km/h) 14 15

Time value per km 0.50 0.87

Total ﬁxed cost per km (C) 0.63 1.32

Table 4: CO

e emission factors, ﬁne particles and land use

per transportation mode. Source: French Environmental

Agency (ADEME).

Cargo bike Vehicles

e emission factors according to the average utilization (ton/km)

10% 0 5.099

20% 0 2.55

30% 0 1.7

40% 0 1.275

50% 0 1.020

60% 0 0.85

70% 0 0.728

80% 0 0.637

90% 0 0.567

100% 0 0.510

Fine particles (g/km) 0 0.01

Land use (m2) 1.77 9.15

Regarding the results obtained in the ﬁrst echelon

(Table 5), it was found that in scenario 2, the num-

ber of vehicles needed to deliver the parcels from the

depot to the satellite in each district of Paris is lower.

Therefore, by using fewer vehicles, CO

e emissions

and ﬁne particles, as well as the ﬁxed cost and land

use are also lower. On the other hand, focusing on

Depot 2, it can be observed that the number of vehi-

cles used and the land use are equal in both scenar-

ios, nevertheless, although CO

e emissions are lower,

these vehicles are traveling a longer distance to de-

liver the parcels to the assigned satellites, which gen-

erates higher ﬁxed costs and more ﬁne particle emis-

sions, both indicators are dependent on the distance

traveled, that in scenario 1 is 1374.79 km and in sce-

nario 2 is 1376.69 km.

Table 5: Results for the ﬁrst echelon: Routing from depots

to satellites.

Scenario Number of vehicles CO

e emissions (kg) Fine particles (g) Fixed cost ( C) Land use (m2)

Depot 1 sc1 24 1167.6 14.9 1966.3 219.6

sc2 23 1004.7 14.5 1912.7 210.5

Depot 2 sc1 20 1063.6 13.7 1814.7 183.0

sc2 20 949.7 13.8 1817.2 183.0

Depot 3 sc1 37 2818.9 35.8 4722.7 338.6

sc2 34 2288.4 32.7 4313.3 311.1

Depot 4 sc1 18 334.8 4.2 552.4 164.7

sc2 19 269.1 3.9 508.6 173.9

Scenario 2 uses cargo bikes for last-mile delivery,

so CO

e and ﬁne particle emissions are zero. Regard-

ing the number of vehicles (Figure 3) and land use,

SIMULTECH 2024 - 14th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

358

for instance, in district 15 scenario 1, the number of

trips or vehicles needed to deliver all the parcels is 12

which occupy 109.8 m2 of public space. However, in

sc2-min there are 35 trips or cargo bikes using 61.9

m2 of land, which means that in District 15 there is

a reduction in land use of 44% and, in relation to the

ﬁxed cost, the cost of scenario 2 is 48% lower than

scenario 1. It is evident that the larger the district and

its demand, the more cargo bikes or trips using them

will be needed, which would cause a quite remark-

able increase of bicycles parked in different areas of

the city such as bike lanes or sidewalks, then it would

be another way to invade public space. Even between

scenario 1 and scenario 2 there is an average reduction

in terms of land use in the second echelon of 32%, and

56% regarding ﬁxed cost (Table 6). As sc2-min and

sc2-max are the ones that considered cargo bikes, we

can say that 3189.8 kg of CO

emissions and 39.7 g of

ﬁne particulate can be saved using this transportation

mode for last-mile delivery either if the delivery time

per node is 2 or 8 minutes. Further studies should

consider the stochasticity of this parameter to make

scenario 2 more realistic.

Figure 3: Number of vehicles or trips per scenario in each

district of Paris (notes: sc2-min refers to the 2min lowest

delivery time, sc2-max to the maximum delivery time of 8

minutes).

About the global results, Table 7 shows the reduc-

tion in CO

e emissions and ﬁne particles when using

scenario 2, because the entire operation of the sec-

ond echelon is executed by cargo bikes that will not

affect the environment with CO

e emissions. There-

fore, these results should be analyzed together with

the other indicators to evaluate their performance so

the company can better understand the impact of ap-

plying the second scenario in its operations. In terms

of ﬁxed cost, it can be noted that in 95% of the dis-

tricts the cost is lower when considering scenario 2,

with an average reduction of 25%. On the other hand,

when analyzing land use, the average reduction in

this indicator between scenario 1 and sc2-min is 17%,

whereas between sc2-min and s2-max there is an in-

Table 6: Second echelon results: Routing from satellites to

clients.

District Scenario CO

e emissions (kg) Fine particles (g) Fixed cost ( C) Land use (m2)

D1 scenario 1 79.1 1.0 127.8 18.3

sc2-min 0.0 0.0 42.0 14.1

sc2-max 0.0 0.0 42.0 38.9

D2 scenario 1 86.9 1.0 131.9 27.5

sc2-min 0.0 0.0 102.0 17.7

sc2-max 0.0 0.0 102.0 51.3

D3 scenario 1 55.8 0.6 84.1 18.3

sc2-min 0.0 0.0 32.7 10.6

sc2-max 0.0 0.0 32.7 35.4

D4 scenario 1 93.5 1.1 138.8 18.3

sc2-min 0.0 0.0 35.8 10.6

sc2-max 0.0 0.0 35.8 31.8

D5 scenario1 88.4 1.1 142.3 27.5

sc2-min 0.0 0.0 66.7 19.4

sc2-max 0.0 0.0 66.7 61.9

D6 scenario 1 84.6 1.0 133.9 27.5

sc2-min 0.0 0.0 55.0 19.4

sc2-max 0.0 0.0 55.0 58.3

D7 scenario 1 115.9 1.6 210.2 45.8

sc2-min 0.0 0.0 83.1 33.6

sc2-max 0.0 0.0 83.1 102.5

D8 scenario 1 207.7 2.4 314.1 54.9

sc2-min 0.0 0.0 115.6 40.7

sc2-max 0.0 0.0 115.6 134.4

D9 scenario 1 152.2 1.8 238.5 45.8

sc2-min 0.0 0.0 81.6 35.4

sc2-max 0.0 0.0 81.6 118.5

D10 scenario 1 169.5 2.1 272.6 45.8

sc2-min 0.0 0.0 95.6 35.4

sc2-max 0.0 0.0 95.6 116.7

D11 scenario 1 192.4 2.3 299.7 64.1

sc2-min 0.0 0.0 135.9 53.0

sc2-max 0.0 0.0 135.9 155.6

D12 scenario 1 221.4 2.7 354.0 64.1

sc2-min 0.0 0.0 192.2 47.7

sc2-max 0.0 0.0 192.2 157.4

D13 scenario 1 191.5 3.1 410.6 82.4

sc2-min 0.0 0.0 184.5 49.5

sc2-max 0.0 0.0 184.5 166.2

D14 scenario 1 198.4 2.4 319.4 45.8

sc2-min 0.0 0.0 135.1 31.8

sc2-max 0.0 0.0 135.1 104.3

D15 scenario 1 286.9 3.4 449.9 109.8

sc2-min 0.0 0.0 233.6 61.9

sc2-max 0.0 0.0 233.6 217.5

D16 scenario 1 252.7 3.4 445.1 82.4

sc2-min 0.0 0.0 218.1 47.7

sc2-max 0.0 0.0 218.1 146.7

D17 scenario 1 236.7 2.9 380.5 73.2

sc2-min 0.0 0.0 176.6 51.3

sc2-max 0.0 0.0 176.6 160.9

D18 scenario 1 169.7 2.1 272.6 45.8

sc2-min 0.0 0.0 127.7 35.4

sc2-max 0.0 0.0 127.7 104.3

D19 scenario 1 133.5 1.6 216.7 36.6

sc2-min 0.0 0.0 100.0 19.4

sc2-max 0.0 0.0 100.0 54.8

D20 scenario 1 173.0 2.1 278.0 54.9

sc2-min 0.0 0.0 142.4 33.6

sc2-max 0.0 0.0 142.4 113.2

crease in land use of 47% because it takes more trips

or more cargo bikes circulating in the city of Paris if

the delivery time is 8 minutes per delivery point.

Two important factors increase the total travel

time. The ﬁrst factor is the speed of the cargo bike,

which is lower than the average speed of the vehi-

cle. Secondly, when adding the delivery time in sce-

nario 2, the increase in travel time is evident because

it is an extra time of 2 to 8 minutes to make what is

known in practice as a delivery in the street. Travel

time in sc2-min and sc2-max generates an increase

in the number of trips because the smaller capacity

of the cargo bikes means that more trips are needed to

cover the total demand of the delivery network. More-

over, working hours are limited to approximately 5.5

hours per day and the driver of the bicycle needs time

to hydrate, have lunch, take some kind of break. In

practice the delivery time is stochastic because the

delivery man does not take the same time in every de-

Adapting Retail Supply Chains for the Race to Sustainable Urban Delivery

359

Table 7: Final results for each scenario.

District Scenario Travel time (h) CO

e emissions (kg) Fine particles emissions (g) Fixed cost ( C) Land use (m2)

D1 scenario 1 14.1 150.9 2.0 259.9 21.8

sc2-min 37.5 95.0 1.4 220.9 32.4

sc2-max 109.6 95.0 1.4 220.9 57.2

D2 scenario 1 20.4 232.9 2.9 376.6 54.9

sc2-min 56.2 150.9 2.2 397.3 45.1

sc2-max 147.6 150.9 2.2 397.3 78.7

D3 scenario 1 12.1 121.9 1.7 223.7 36.6

sc2-min 34.7 100.2 1.4 218.6 28.9

sc2-max 142.0 100.2 1.4 218.6 53.7

D4 scenario 1 10.3 117.7 1.4 190.6 36.6

sc2-min 26.2 35.6 0.5 101.6 28.9

sc2-max 83.5 35.6 0.5 101.6 50.1

D5 scenario1 13.9 159.2 1.9 256.6 54.9

sc2-min 46.6 59.5 0.9 181.2 46.9

sc2-max 155.2 59.5 0.9 181.2 89.3

D6 scenario 1 14.2 165.6 2.0 263.1 54.9

sc2-min 42.4 64.4 0.9 175.4 46.9

sc2-max 155.3 64.4 0.9 175.4 85.8

D7 scenario 1 37.0 381.5 5.2 684.2 91.5

sc2-min 95.0 211.3 3.1 491.4 70.2

sc2-max 289.8 211.3 3.1 491.4 139.1

D8 scenario 1 52.82 613.96 7.4 976.2 109.8

sc2-min 129.9 336.3 4.9 758.2 86.4

sc2-max 370.2 336.3 4.9 758.2 180.1

D9 scenario 1 32.9 362.7 4.6 607.8 91.5

sc2-min 100.2 177.5 2.5 418.2 72.0

sc2-max 324.1 177.5 2.5 418.2 155.1

D10 scenario 1 42.3 472.7 5.9 782.6 91.5

sc2-min 107.2 239.9 3.4 549.3 72.0

sc2-max 330.2 239.9 3.4 549.3 153.3

D11 scenario 1 42.6 484.8 6.0 786.7 128.1

sc2-min 133.1 245.4 3.6 606.4 107.9

sc2-max 418.3 245.4 3.6 606.4 210.5

D12 scenario 1 46.7 529.5 6.5 862.3 128.1

sc2-min 136.0 249.6 3.5 660.6 102.6

sc2-max 412.4 249.6 3.5 660.6 212.3

D13 scenario 1 63.1 659.4 8.8 1167.0 155.6

sc2-min 152.0 382.4 5.5 909.9 122.7

sc2-max 440.2 382.4 5.5 909.9 239.4

D14 scenario 1 41.5 476.1 5.8 767.6 91.5

sc2-min 100.9 237.6 3.5 594.6 77.6

sc2-max 291.5 237.6 3.5 594.6 150.1

D15 scenario 1 100.1 1116.7 14.0 1849.4 219.6

sc2-min 211.7 659.4 9.3 1459.2 141.0

sc2-max 585.7 659.4 9.3 1459.2 257.7

D16 scenario 1 86.6 923.2 12.1 1599.6 164.7

sc2-min 162.0 557.2 8.1 1287.4 120.9

sc2-max 417.4 557.2 8.1 1287.4 219.9

D17 scenario 1 74.5 845.8 10.4 1377.0 146.4

sc2-min 160.0 495.6 7.0 1098.8 115.3

sc2-max 444.5 495.6 7.0 1098.8 224.9

D18 scenario 1 22.8 261.6 3.2 421.0 91.5

sc2-min 80.9 58.2 0.8 237.9 81.1

sc2-max 266.7 58.2 0.8 237.9 150.1

D19 scenario 1 23.7 258.6 3.3 438.1 73.2

sc2-min 50.8 104.4 1.5 301.5 46.9

sc2-max 141.2 104.4 1.5 301.5 82.3

D20 scenario 1 20.9 239.9 2.9 386.7 109.8

sc2-min 84.9 51.4 0.7 240.1 88.5

sc2-max 280.0 51.4 0.7 240.1 168.1

livery point, that is why one scenario is considered

with more optimistic times than the other, to analyze

these differences. Finally, the increase in total travel

time between Scenario 1 and Scenario 2 is approxi-

mately 62%, while the increase between sc2-min and

sc2-max is 67%.

6 CONCLUSIONS AND

PERSPECTIVES

The retail industry has evolved with the increasing

expansion of e-commerce. Disruptive unexpected

events, such as the COVID-19 pandemic, have rein-

forced this global trend, so that supply chain and lo-

gistics operations have been redesigned. Since last-

mile supply chains are the core of e-commerce deliv-

ery operations, effective operational behavior is key to

allow e-commerce business to continue growing. In

addition, the conﬁguration of last-mile supply chains

has evolved to nowadays require the design of two-

echelon urban delivery networks, in which urban con-

solidation centers (UCC) are key nodes between the

main supplier and the ﬁnal customer. This network

structure is known in the literature as a two-echelon

supply chain.

This paper analyzed the problem of designing de-

livery routes in such two-echelon distribution sys-

tems. Because of its complexity to be solved when

dealing with large sized datasets, this paper evalu-

ated the implementation of an approximation algo-

rithm based on the well-known NN routing heuristic.

A set of numerical experiments were carried out on

datasets taken from a real-life French company deliv-

ering products to more than 90,000 points in Paris.

A heterogeneous ﬂeet of delivery vehicles was evalu-

ated, using vehicles and cargo bikes in the ﬁrst and

the second echelon, respectively. Different opera-

tional scenarios were evaluated, and the performance

of the system was assessed. Results show that the

consideration of mobile storage instead of conven-

tional satellites can generate savings not only in emis-

sions, ﬁxed costs and land use, but also regarding in-

vestment cost of locating satellite facilities within the

city because transshipment activities between the de-

livery van and the cargo bike could be performed in

smaller areas such as a parking lots or public space

areas designated only for these purposes, while satel-

lites in scenario 1 could provide parcel storage ser-

vice, which leads to additional costs not considered

in this study. Moreover, the inclusion of eco-friendly

modes of transportation like cargo bikes has a lot of

potential to reduce some of the externalities caused

by the usual urban parcel delivery systems that have

conventional vans. Furthermore, although the consid-

eration of scenario 2 increases the invasion of public

space in inner-city areas, it does guarantee important

savings in CO

e and ﬁne particulate emissions and,

ﬁxed costs are lower. However, given the advantages

of using cargo bikes in the second scenario, a mix be-

tween electric vans and cargo bikes, taking advantage

of the beneﬁts of each transportation mode, could be

considered in future research.

Based on the outcomes of this research, several

opportunities for further research are open. The re-

search on 2E-VRP and its variants has a lot of oppor-

tunities for example, the consideration of additional

constraints, like load or route balancing among ve-

hicles, stochastic delivery times, heterogeneous ﬂeet

in the same echelon (electric vehicles, electric cargo

bikes, etc.), the design of distribution network con-

ﬁgurations with multimodal transportation for parcel

distribution systems in big cities, among other chal-

lenges. Another line for future research is the design

of other solution procedures, especially to deal with

very large amounts of delivery points. Finally, since

the proposed solution approach is used to generate a

solution to the problem, another research avenue will

address the design of improvement heuristics to ob-

tain better results in terms of routes.

SIMULTECH 2024 - 14th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

360

ACKNOWLEDGEMENTS

The work of the ﬁrst author was carried out un-

der a post-graduate scholarship awarded by Univer-

sidad de La Sabana, Colombia and Kedge Business

School, France. This work was also supported by

research grants INGPHD-52-2022 and INGPHD-10-

2019 from Universidad de La Sabana, Colombia.

REFERENCES

Anderluh, A., Hemmelmayr, V., and Nolz, P. (2019). Chap-

ter 8 - sustainable logistics with cargo bikes—methods

and applications. In Faulin, J., Grasman, S. E., Juan,

A. A., and Hirsch, P., editors, Sustainable Transporta-

tion and Smart Logistics, pages 207–232. Elsevier.

Bendoly, E., Craig, N., and DeHoratius, N. (2018). Consis-

tency and recovery in retail supply chains. Journal of

Business Logistics, 39(1):26–37.

Browne, M., Sweet, M., Woodburn, A., and Allen, J.

(2005). Urban freight consolidation centres ﬁnal re-

port. Technical report, University of Westminster.

Buijs, P., Alvarez, J., Veenstra, M., and Roodbergen, K.

(2016). Improved collaborative transport planning at

dutch logistics service provider fritom. Interfaces,

46(2):119–132.

Caggiani, L., Colovic, A., Prencipe, L. P., and Ottomanelli,

M. (2021). A green logistics solution for last-mile de-

liveries considering e-vans and e-cargo bikes. Trans-

portation Research Procedia, 52:75–82.

Carlsson, J. and Song, S. (2018). Coordinated logistics

with a truck and a drone. Management Science,

64(9):4052–4069.

Cattaruzza, D., Absi, N., Feillet, D., and Gonz

alez-Feliu,

J. (2017). Vehicle routing problems for city logis-

tics. EURO Journal on Transportation and Logistics,

6(1):51–79.

Crainic, T., Perboli, G., Mancini, S., and Tadei, R. (2010).

Two-echelon vehicle routing problem: A satellite lo-

cation analysis. Procedia - Social and Behavioral Sci-

ences, 2(3):5944–5955.

Gonzalez-Feliu, J., Perboli, G., Tadei, R., and Vigo, D.

(2008). The two-echelon capacitated vehicle routing

problem.

Jacobsen, S. and Madsen, O. (1980). A comparative study

of heuristics for a two-level routing-location problem.

European Journal of Operational Research, 5(6):378–

387.

Juan, A., Faulin, J., Grasman, S., Rabe, M., and Figueira, G.

(2015). A review of simheuristics: Extending meta-

heuristics to deal with stochastic combinatorial op-

timization problems. Operations Research Perspec-

tives, 2:62–72.

Kull, T., Boyer, K., and Calantone, R. (2007). Last-mile

supply chain efﬁciency: an analysis of learning curves

in online ordering. International Journal of Opera-

tions & Production Management, 27(4):409–434.

Lone, S., Harboul, N., and Weltevreden, J. (2021). Eu-

ropean e-commerce report. Technical report, E-

Commerce Europe.

Marrekchi, E., Besbes, W., Dhouib, D., and Demir, E.

(2021). A review of recent advances in the oper-

ations research literature on the green routing prob-

lem and its variants. Annals of Operations Research,

304(1):529–574.

Meza-Peralta, K., Gonzalez-Feliu, J., Montoya-Torres, J.,

and Khodadad-Saryazdi, A. (2020). A uniﬁed typol-

ogy of urban logistics spaces as interfaces for freight

transport. Supply Chain Forum: An International

Journal, 21(4):274–289.

Montoya-Torres, J., Mu

noz-Villamizar, A., and Mejia-

Argueta, C. (2021). Mapping research in logistics and

supply chain management during covid-19 pandemic.

International Journal of Logistics Research and Ap-

plications, 26(4):421–441.

Nguyen, D., de Leeuw, S., Dullaert, W., and Foubert, B.

(2019). What is the right delivery option for you? con-

sumer preferences for delivery attributes in online re-

tailing. Journal of Business Logistics, 40(4):299–321.

Patier, D. and Browne, M. (2010). A methodology for the

evaluation of urban logistics innovations. Procedia -

Social and Behavioral Sciences, 2(3):6229–6241.

Ramirez-Villamil, A., Jaegler, A., and Montoya-Torres, J.

(2022). Sustainable local pickup and delivery: The

case of paris. Research in Transportation Business &

Management, 45:100692.

Reed, S., Campbell, A., and Thomas, B. (2022). The value

of autonomous vehicles for last-mile deliveries in ur-

ban environments. Management Science, 68(1):280–

299.

Sandhaus, G. (2019). Trends in E-commerce, Logistics and

Supply Chain Management, pages 593–610. Springer

International Publishing, Cham.

Silva, V., Amaral, A., and Fontes, T. (2023). Towards sus-

tainable last-mile logistics: A decision-making model

for complex urban contexts. Sustainable Cities and

Society, 96:104665.

Soeffker, N., Ulmer, M. W., and Mattfeld, D. C. (2022).

Stochastic dynamic vehicle routing in the light of pre-

scriptive analytics: A review. European Journal of

Operational Research, 298(3):801–820.

Stamadianos, T., Kyriakakis, N., Marinaki, M., and Mari-

nakis, Y. (2023). Routing problems with electric and

autonomous vehicles: Review and potential for future

research. Operations Research Forum, 4(2).

Stroh, A., Erera, A., and Toriello, A. (2022). Tactical design

of same-day delivery systems. Management Science,

68(5):3444–3463.

Taiwo, O., Josiah, O., Taiwo, A., Dkhrullahi, S., and Sade,

O. K. (2013). Implementation of heuristics for solv-

ing travelling salesman problem using nearest neigh-

bor insertion approaches. International Journal of Ad-

vanced Research, 1(3):139–155.

Vega-Mej

ıa, C., Montoya-Torres, J., and Islam, S. (2019).

Consideration of triple bottom line objectives for sus-

tainability in the optimization of vehicle routing and

loading operations: a systematic literature review. An-

nals of Operations Research, 273(1):311–375.

Adapting Retail Supply Chains for the Race to Sustainable Urban Delivery

361

Zhou, L., Baldacci, R., Vigo, D., and Wang, X. (2018).

A multi-depot two-echelon vehicle routing problem

with delivery options arising in the last mile distri-

bution. European Journal of Operational Research,

265(2):765–778.

SIMULTECH 2024 - 14th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

362