Available online at www.sciencedirect.

com

ScienceDirect
Available online at www.sciencedirect.com
Available online at www.sciencedirect.com
ScienceDirect
Transportation Research Procedia 00 (2023) 000–000
www.elsevier.com/locate/procedia
ScienceDirect
Transportation Research Procedia 00 (2023) 000–000
Transportation Research Procedia 78 (2024) 32–39
www.elsevier.com/locate/procedia

25th Euro Working Group on Transportation Meeting (EWGT 2023)

Integration of Google
25th Euro MapsonAPI
Working Group with mathematical
Transportation modeling
Meeting (EWGT 2023) for
solving the Real-Time VRP
Integration of Google Maps API with mathematical modeling for
Andres Muñoz-Villamizara,*, Javier Faulinb, Lorena Reyes-Rubianoa
solving the Real-Time VRP
Rafael Henriquez-Machadoa, Elyn Solano-Charrisa
Andres Muñoz-Villamizar
International
a
a,
*, JavierSciences,
School of Economic and Administrative Faulin b
, Lorena
Universidad Reyes-Rubiano
de La Sabana, Chía, Colombia
a

Institute of Rafael
b
Henriquez-Machado
Smart Cities. , Elyn Solano-Charrisa
a Public University of Navarra, Pamplona,
Department of Statistics and OR. Spain

a
International School of Economic and Administrative Sciences, Universidad de La Sabana, Chía, Colombia
b
Institute of Smart Cities. Department of Statistics and OR. Public University of Navarra, Pamplona, Spain

Abstract:
Freight transportation is the backbone of urban economies and plays a critical role in the smooth functioning of cities.
As such, devising efficient methods for freight transportation planning is of paramount importance. One of the most
Abstract:
crucial aspects affecting the efficiency of freight transport is the variability in travel speeds, impacted by factors such
Freight
as traffictransportation is the backbone
congestion. While traditionalofapproaches
urban economies andon
often rely plays
GPSa technologies
critical role inand
theassociated
smooth functioning of cities.
routing services—
As such, devising efficient methods for freight transportation planning is of paramount importance.
which can be expensive—for planning, these methods also necessitate frequent re-optimization due to ever-changing One of the most
crucial
traffic conditions. To address these challenges, we introduce a novel solution that integrates real-time traffic datasuch
aspects affecting the efficiency of freight transport is the variability in travel speeds, impacted by factors into
as traffic
daily congestion.
vehicle While traditional
route planning. approaches
Specifically, our method often rely on GPS
incorporates technologies
Google Maps APIand forassociated routing services—
traffic congestion estimation
which can bea expensive—for
and utilizes Mixed Integer Linearplanning, these methods
Programming also necessitate
(MILP) frequent re-optimization
model to determine optimal routes for dueantoentire
ever-changing
day. We
tested our methodology in a major U.S. city and found that it outperforms conventional approaches by up todata
traffic conditions. To address these challenges, we introduce a novel solution that integrates real-time traffic 18%into
in
daily vehicle route planning. Specifically, our method incorporates
terms of routing time, underscoring its practical relevance and efficiency. Google Maps API for traffic congestion estimation
and utilizes a Mixed Integer Linear Programming (MILP) model to determine optimal routes for an entire day. We
© 2024our
tested Themethodology
Authors. Published by ELSEVIER
in a Problem;
major U.S. B.V.
cityLogistics;
and found that it Maps
outperforms conventional approaches by up to 18% in
Keywords:
This is an Vehicle
open accessRouting
article under the CC City
BY-NC-ND Google
license API; Case Study
(https://creativecommons.org/licenses/by-nc-nd/4.0)
terms of routing time, underscoring its practical relevance and efficiency.
Peer-review under responsibility of the scientific committee of the 25th Euro Working Group on Transportation Meeting (EWGT 2023)
1. Introduction
Keywords: Vehicle Routing Problem; City Logistics; Google Maps API; Case Study
Freight transportation is a cornerstone of urban logistics, and the efficiency of its planning and execution is
1. Introduction
paramount for smooth city operations. One of the most critical factors affecting this efficiency is the variability in
travel speeds, influenced by changing traffic conditions throughout the day. While Information and Communication
Freight transportation is a cornerstone of urban logistics, and the efficiency of its planning and execution is
paramount for smooth city operations. One of the most critical factors affecting this efficiency is the variability in
travel speeds, influenced by changing traffic conditions throughout the day. While Information and Communication

* Corresponding author. Tel.: +57 861 5555

E-mail address: [email protected]

* Corresponding
2352-1465 © 2023 author. Tel.: +57
The Authors. 861 5555
Published by Elsevier BV.
ThisE-mail address:
is an open [email protected]
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 25th Euro Working Group on Transportation Meeting (EWGT
2023)
2352-1465 © 2023 The Authors. Published by Elsevier BV.
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 25th Euro Working Group on Transportation Meeting (EWGT
2352-1465 © 2024 The Authors. Published by ELSEVIER B.V.
2023)
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 25th Euro Working Group on Transportation Meeting (EWGT 2023)
10.1016/j.trpro.2024.02.005
 Andres Muñoz-Villamizar et al. / Transportation Research Procedia 78 (2024) 32–39 33
2 Munoz-Villamizar et al./ Transportation Research Procedia 00 (2023) 000–000

Technologies (ICT) offer a robust platform for real-time traffic data, their efficacy in enhancing freight transportation
largely depends on the ability to design and adapt optimal routes in real-time (Liao & Hu, 2011).
Traditional approaches often rely on recalculating the shortest path after each node is visited, providing a reasonable
yet often inaccurate estimate of the total travel time (Gentile et al., 2020). Moreover, these approaches often adapt
static methodologies that do not consider real-time variations in city traffic (Liao & Hu, 2011). In light of these
limitations, this paper addresses the central research question: How can travel times in urban settings be minimized in
real-time?
To answer this question, we propose a novel solution that leverages Google Maps API for real-time traffic
congestion estimation. This data feeds into a Mixed Integer Linear Programming (MILP) model to optimize vehicle
routes under actual traffic conditions. Our methodology was validated using real-world data from Boston,
Massachusetts, across 300 separate instances representing different travel times and three distinct time slots.
Comparative analysis with traditional methods demonstrates that our approach significantly reduces travel times,
outperforming conventional strategies.
The contributions of this research extend beyond algorithmic improvements to offer valuable insights for decision-
makers in freight transport. Specifically, our work integrates Google Maps API with MILP models, allowing for real-
time adaptation of vehicle routes and offering an empirical analysis of how different traffic conditions affect
transportation efficiency at various times of the day.
The remainder of this paper is structured as follows: Section 2 reviews the relevant literature, Section 3 outlines
our solution approach, Section 4 presents a case study based on real-world data, and Section 5 concludes the paper
and suggests directions for future research.

2. Literature Review

In this section, we survey existing literature on Real-Time Vehicle Routing Problems (RT-VRP) and the benefits
and challenges of utilizing Application Programming Interfaces (APIs) like Google Maps for route optimization.

2.1. Real-Time VRP

RT-VRPs find applications in both commercial goods transport and public transportation. The aim of traditional
Vehicle Routing Problems (VRP) is to devise an optimal route and schedule that meets customer demands while
minimizing operational costs. However, in real-world scenarios, static VRPs face challenges due to the dynamic nature
of traffic conditions (Cardoso et al., 2015). The early 2000s saw the first review of RT-VRP by Ghiani et al. (2003),
emphasizing the need for real-time data, particularly as the internet was in its nascent stage.
Dynamic Vehicle Routing Problems (DVRP) account for variables that change over time, including the
environmental impact of traffic and transportation (Zhang et al., 2021; Alvarez et al., 2018). There are several variants
of VRP, including VRP with time windows (VRPTW), pickup and delivery (VRPPD), stochastic demands (VRPSD),
and others (Jia et al., 2018). These variations, like their traditional counterpart, rely heavily on data for decision-
making.
RT-VRP adds a layer of complexity by accounting for real-time information on weather and traffic conditions,
aiding in avoiding road obstacles and enhancing communication between dispatchers and drivers (Okhrin & Richter,
2008; Lianxi, 2012). Various tools and technologies have been developed to meet these demands, including meta-
heuristic approaches for minimum-cost planning in dynamic settings (Jia et al., 2018).

2.2. Tools and Technology

Among the technologies employed in solving RT-VRPs, Google APIs stand out. Developed by Google, these
Application Programming Interfaces (APIs) facilitate communication with Google Services and can be integrated with
other interfaces. They serve as platforms for real-time distance and transportation information, and also offer
comprehensive cloud services that handle diverse data and support IoT protocols and endpoints (Google Maps
Platforms, 2021; Lim et al., 2021). In addition to Google APIs, Genetic Programming (GP) is another important tool.
Originating from the field of artificial intelligence, GP operates by evolving a population of programs, often initially
34 Andres Muñoz-Villamizar et al. / Transportation Research Procedia 78 (2024) 32–39
Munoz-Villamizar et al. / Transportation Research Procedia 00 (2019) 000–000 3

random, to complete specific tasks. This is done through mechanisms similar to natural genetic processes. GP has
proven effective in resolving complex combinatorial optimization challenges, including those with real-time variables
such as travel time and cost (Jacobsen-Grocott et al., 2017). The main challenges GP faces include designing effective
resource policies and obtaining tasks in a linear sequence without compromising quality (Jia et al., 2018).
Dynamic programming is yet another method employed for real-time vehicle routing. This approach leverages
semi-infinite linear programming to solve Dynamic Vehicle Routing Problems (DVRP). By working with variables
that define intermediate times, dynamic programming aims to minimize the total time taken when compared to static
VRP models (Yu & Yang, 2019). Agent-based simulation models also offer an alternative approach to solving DVRPs.
These models identify and simulate the interactions among various actors involved, providing a more contextual
solution to the problem (Alvarez et al., 2018).
The Adaptive Neighborhood Search (ANS) heuristic is another methodology worth noting. It is particularly
effective for leveraging the specific structure of time intervals. ANS assesses the feasibility of incorporating a new
order or purchase within a given time frame, doing so in a computationally efficient manner (Hungerländer et al.,
2017). Finally, Neural Network Models are emerging as a popular tool as well, especially for more complex versions
of stochastic VRPs. Unlike other models, they do not require the pre-calculation of a distance matrix. They are
designed to scale effectively as the problem size increases, thus offering superior performance and competitive
solution times (Nazari et al., 2018).

3. Proposed Approach

3.1. General Overview

The conventional methods of routing optimization often employ proprietary software and technologies that
continuously update real-time traffic data, leading to constant recalculations of routing plans. In contrast, our approach
aims to harness freely available technologies to predict traffic behavior on a pre-defined network and establish a static
routing plan. This static plan aims to outperform traditional real-time optimization methods in terms of both cost-
efficiency and stability of the initially planned route. Figure 1 provides a comparative overview of the traditional and
our proposed models, emphasizing the replicability of our methodology and its applicability to case-study research
(Yin, 2003).

Figure 1. Generic comparison for traditional and the proposed approach for solving Real-Time VRP
 Andres Muñoz-Villamizar et al. / Transportation Research Procedia 78 (2024) 32–39 35
4 Munoz-Villamizar et al./ Transportation Research Procedia 00 (2023) 000–000

The first step in our methodology is to create a mathematical structure to represent the network, the mobility of the
vehicle fleet, and the locations of delivery nodes that must be visited during the routing operation. The second step
involves the estimation of speed and distance across the network within a specific planning horizon. To this end, we
employ the Google Maps API, which covers an extensive network of roads in over 200 countries and territories. This
API not only calculates current travel times based on real-time data but also predicts future travel times using historical
data (Google Cloud, 2021). Thus, the API enables us to obtain round-trip travel distances and times between each pair
of nodes for discrete future time slots (e.g., 8 am to 9 am, 9 am to 10 am, etc.). The third step utilizes the arrays
generated in the second step. Specifically, a Mixed Integer Linear Programming (MILP) model is employed to tackle
the Real-time Vehicle Routing Problem (VRP). The model's notation and details will be elaborated in subsection 3.2.
Finally, in the fourth step, the optimal solution generated by the MILP model serves as the definitive routing plan.
One of the unique advantages of our approach is that, owing to the prior traffic condition estimations made possible
by the Google Maps API, there is no need for frequent updates or recalculations.

3.2. MILP model for Real-Time VRP

We formalize the Real-Time Vehicle Routing Problem (RT-VRP) as a directed graph 𝐺𝐺𝐺𝐺 = (𝑉𝑉𝑉𝑉, 𝐴𝐴𝐴𝐴), where 𝑉𝑉𝑉𝑉 = 𝑁𝑁𝑁𝑁 ∪
{0} and N = {1, . . ., n} represents n customers, with node 0 being the depot. The planning horizon is denoted by T=
{1, …, t}. Each customer is labeled with i and j, such that 𝑖𝑖𝑖𝑖 ∈ 𝑁𝑁𝑁𝑁 . For each time slot t, the set of arcs 𝐴𝐴𝐴𝐴𝑡𝑡𝑡𝑡 =
{(𝑖𝑖𝑖𝑖, 𝑗𝑗𝑗𝑗): 𝑖𝑖𝑖𝑖, 𝑗𝑗𝑗𝑗 ∈ 𝑉𝑉𝑉𝑉, 𝑖𝑖𝑖𝑖 ≠ 𝑗𝑗𝑗𝑗} signifies the travel time between each pair of nodes. This time is quantified as 𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡𝑖𝑖𝑖𝑖 in minutes.
To discretize the time slots within the routes, we use a vector 𝑑𝑑𝑑𝑑𝑡𝑡𝑡𝑡 defined as 𝑑𝑑𝑑𝑑1 = 1, 𝑑𝑑𝑑𝑑2 = 2, … 𝑑𝑑𝑑𝑑𝑡𝑡𝑡𝑡 = 𝑡𝑡𝑡𝑡. Lastly, M
represents a large positive number.
Furthermore, binary variables are defined as: 𝑌𝑌𝑌𝑌𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡𝑖𝑖𝑖𝑖 = 1 if customer i precedes j at time slot t, otherwise it equals 0.
Besides, non-negative variable 𝑋𝑋𝑋𝑋𝑖𝑖𝑖𝑖 is defined as the arrival time [minutes] to customer i. Additionally, auxiliary
variables 𝑢𝑢𝑢𝑢𝑖𝑖𝑖𝑖 are used to prevent subtours. The following notation and model are used:

min 𝑍𝑍𝑍𝑍 = � � � 𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡𝑖𝑖𝑖𝑖 ∙ 𝑌𝑌𝑌𝑌𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡𝑖𝑖𝑖𝑖

𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑡𝑡𝑡𝑡 (1)

𝑡𝑡𝑡𝑡
� � 𝑌𝑌𝑌𝑌0𝑖𝑖𝑖𝑖 =1 (2)
𝑖𝑖𝑖𝑖 𝑡𝑡𝑡𝑡

� � 𝑌𝑌𝑌𝑌𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡0 = 1 (3)
𝑖𝑖𝑖𝑖 𝑡𝑡𝑡𝑡
� � 𝑌𝑌𝑌𝑌𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡𝑖𝑖𝑖𝑖 = 1; ∀𝑗𝑗𝑗𝑗 (4)
𝑖𝑖𝑖𝑖 𝑡𝑡𝑡𝑡

� � 𝑌𝑌𝑌𝑌𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡𝑖𝑖𝑖𝑖 = � � 𝑌𝑌𝑌𝑌𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡 ; ∀𝑗𝑗𝑗𝑗 > 0 (5)

𝑖𝑖𝑖𝑖 𝑡𝑡𝑡𝑡 𝑖𝑖𝑖𝑖 𝑡𝑡𝑡𝑡
𝑡𝑡𝑡𝑡 𝑡𝑡𝑡𝑡
�1 − 𝑌𝑌𝑌𝑌0𝑖𝑖𝑖𝑖 � ∙ 𝑀𝑀𝑀𝑀 + 𝑋𝑋𝑋𝑋𝑖𝑖𝑖𝑖 ≥ 𝑐𝑐𝑐𝑐0𝑖𝑖𝑖𝑖 ; ∀𝑗𝑗𝑗𝑗, ∀𝑡𝑡𝑡𝑡 (6)

�1 − 𝑌𝑌𝑌𝑌𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡𝑖𝑖𝑖𝑖 � ∙ 𝑀𝑀𝑀𝑀 + �𝑋𝑋𝑋𝑋𝑖𝑖𝑖𝑖 − 𝑋𝑋𝑋𝑋𝑖𝑖𝑖𝑖 � ≥ 𝑐𝑐𝑐𝑐𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡𝑖𝑖𝑖𝑖 ; ∀𝑖𝑖𝑖𝑖

(7)
> 0, ∀𝑗𝑗𝑗𝑗, ∀𝑡𝑡𝑡𝑡

𝑌𝑌𝑌𝑌𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡𝑖𝑖𝑖𝑖 ∙ 𝑑𝑑𝑑𝑑𝑡𝑡𝑡𝑡 + �1 − 𝑌𝑌𝑌𝑌0𝑖𝑖𝑖𝑖

𝑡𝑡𝑡𝑡
� ∙ 𝑀𝑀𝑀𝑀 ≥ 𝑋𝑋𝑋𝑋𝑖𝑖𝑖𝑖 ; ∀𝑖𝑖𝑖𝑖, ∀𝑗𝑗𝑗𝑗, ∀𝑡𝑡𝑡𝑡 (8)

𝑢𝑢𝑢𝑢𝑖𝑖𝑖𝑖 − 𝑢𝑢𝑢𝑢𝑖𝑖𝑖𝑖 + 𝑛𝑛𝑛𝑛 ∙ 𝑌𝑌𝑌𝑌𝑖𝑖𝑖𝑖𝑡𝑡𝑡𝑡𝑖𝑖𝑖𝑖 ≤ 𝑛𝑛𝑛𝑛 − 1; ∀𝑖𝑖𝑖𝑖 > 0, ∀𝑗𝑗𝑗𝑗 > 0, ∀𝑡𝑡𝑡𝑡 (9)

The objective function (1) minimizes the total routing time. Constraints (2) and (3) ensure that route starts and ends
at the depot, respectively. Constraint (4) enforces that all customers, except the depot, are visited exactly once.
Moreover, routing continuity is represented by Constraint (5). Constraints (6) and (7) calculate the arrival time for the
36 Andres Muñoz-Villamizar et al. / Transportation Research Procedia 78 (2024) 32–39
Munoz-Villamizar et al. / Transportation Research Procedia 00 (2019) 000–000 5

first and the rest of the customers, respectively. Likewise, constraint (8) establishes the slot time in which each
customer is visited. Finally, constraint (9) Eliminates the possibility of subtours.

4. Results and Analysis

We applied the methodology outlined in Section 3 to a real-world setting, focusing on the city of Boston,
Massachusetts. Recognizing that the RT-VRP is a specialized form of the known NP-hard problem VRP, constraints
were imposed on both the planning horizon t and the number of customers i or j to be visited. Initial steps involved an
exploratory analysis to better comprehend the traffic dynamics within the city, and thus, to define the scope of the
experiment. Using Google Maps API, we collected data on as many as 12,000 route distances and times across
different zip codes in Boston. These data spanned a 24-hour period to allow for a nuanced understanding of daily
traffic patterns. The average speed in the city ranged from 9.41 km/h to 33.12 km/h, with an overall average speed of
approximately 21.27 km/h. As expected, areas near the city center showed higher traffic congestion and slower
average speeds. For the study, we earmarked 20 customers to be visited. It is worth noting that the choice of 20
customers was deliberately made to bypass the need for incorporating capacity constraints, which are typically not an
issue in last-mile delivery scenarios where vans often make more than 20 stops without exceeding volume capacity.
Instead, our focus was solely on real-time impact. To offer managerial insights, we established three different time
slots, or scenarios: Scenario A (6 am to 10 am), Scenario B (10 am to 2 pm), and Scenario C (3 pm to 7 pm). These
scenarios were designed to capture the variability of traffic conditions throughout the day.

Figure 2. Exploratory study in Boston, US.

6 Munoz-Villamizar et al./ Transportation Research Procedia 00 (2023) 000–000
Andres Muñoz-Villamizar et al. / Transportation Research Procedia 78 (2024) 32–39 37

To test the robustness of our model, we created 300 different instances for each scenario, varying customer
locations, travel distances, time of the route, and daytime conditions. Each of these instances was solved using the
MILP model presented in Section 3.2, employing Gurobi version 9.1. Remarkably, for each scenario, the optimal
solution was obtained in less than 1 minute. The test bed comprised an Intel 8-Core i9 PC with a 2.3 GHz CPU and
16 GB of RAM.
The results are summarized in Table 1 and provide a compelling argument for the efficacy of our RT-VRP approach
compared to traditional VRP methods. On average, our model improved the total route time by 6.8%, 7.7%, and 17.6%
in Scenarios A, B, and C, respectively. Figure 4 offers a more detailed analysis. Specifically, Figure 4a elucidates the
relationship between the total route time and the savings achieved through our methodology. In Scenarios A and C,
we observed diminishing returns on savings as route time increased. Conversely, Scenario B showed consistent
savings even with extended route times. Figure 4b explores a similar analysis for route distance. Figure 4c presents a
somewhat contrasting observation; it revealed that route time and distance are not perfectly correlated. A shorter travel
time does not inherently guarantee a shorter distance traveled. This divergence is largely because travel time can be
influenced by various factors like road conditions and traffic lights, which don't necessarily change the distance.

Table 1. Average results per Scenario.

Average Route time (min) Absolute Improvement
Scenario
Traditional Approach Proposed Approach (min)
A 171.56 159.95 11.61
B 176.18 162.62 13.56
C 180.20 148.46 31.74

3.a 3.b
38 Andres Muñoz-Villamizar et al. / Transportation Research Procedia 78 (2024) 32–39
Munoz-Villamizar et al. / Transportation Research Procedia 00 (2019) 000–000 7

3.c

Figure 3. Route Time versus Time Improvement and Distance for the Proposed Approach

5. Conclusion and further research

This article presents a novel solution method for daily vehicle routing that takes into account traffic behavior.
Utilizing Google Maps API and a Mixed Integer Linear Programming model, we achieve optimal routing without the
need for GPS technologies in vehicles or constant route updates. Our results underline the importance of real-time
data for the real-time optimization of vehicle routes. Key conclusions from our case study can be summarized as
follows:
a) The RT-VRP model significantly improves time performance, especially when external factors that delay
initial planned routes are more pronounced.
b) As traffic data becomes increasingly affordable and accessible, solving real-world traffic issues becomes
more effective through the use of such real-time data sources.
In terms of future research, several avenues present themselves for exploration. First, the scope of real-time data
could be expanded to include sustainability metrics. This would involve capturing and analyzing data related to traffic
noise, air pollution, infrastructure wear, and traffic harmonization. Properly assessed, these metrics have the potential
to make mobility management a more informed and responsible activity, particularly for human drivers. Second, as
autonomous vehicles become more prevalent, the role of real-time data in vehicle routing will become increasingly
important, marking this as a vital area for subsequent studies. Additionally, the issue of data storage cannot be
overlooked; more research is needed to determine optimal storage solutions capable of handling the large amounts of
real-time data generated by this approach. Lastly, while the current study deliberately focuses on time constraints to
the exclusion of vehicle capacity, future work could explore the integration of capacity constraints into the RT-VRP
model for a more comprehensive approach to vehicle routing

References
Alvarez, P., Lerga, I., Serrano-Hernandez, A., & Faulin, J. (2018). The impact of traffic congestion when optimising delivery routes in real time.
A case study in Spain. Int. J. Logist. Res. Appl., 21(5), 529–541. https://doi.org/10.1080/13675567.2018.1457634
Cardoso, P. J. S., Schütz, G., Mazayev, A., Ey, E., & Corrêa, T. (2015). A Solution for a Real-time Stochastic Capacitated Vehicle Routing Problem
with Time Windows. Procedia Comput. Sci., 51, 2227–2236. https://doi.org/10.1016/j.procs.2015.05.501
Ferrucci, F., Bock, S., & Gendreau, M. (2013). A pro-active real-time control approach for dynamic vehicle routing problems dealing with the
delivery of urgent goods. Eur. J. Oper. Res., 225(1), 130–141. https://doi.org/10.1016/j.ejor.2012.09.016
 Andres Muñoz-Villamizar et al. / Transportation Research Procedia 78 (2024) 32–39 39
8 Munoz-Villamizar et al./ Transportation Research Procedia 00 (2023) 000–000

Gentile, G., Meschini, L., Tiddi, D., Fiorani, A., & Attanasi, A. (2020). Fast estimation of point-to-point travel times for real-time vehicle routing.
Transp. Res. Proc., 47, 401–408. https://doi.org/10.1016/j.trpro.2020.03.115
Ghiani, G., Guerriero, F., Laporte, G., & Musmanno, R. (2003). Real-time vehicle routing: Solution concepts, algorithms and parallel computing
strategies. Eur. J. Oper. Res., 151(1), 1–11. https://doi.org/10.1016/S0377-2217(02)00915-3
Google Cloud. (2021). Google Maps Platform. https://cloud.google.com/maps-platform/routes (Accessed may, 2022).
Google Maps Platform. (2021) Available online: https://developers.google.com/maps/documentation/distance-matrix/overview (Accessed may,
2022).
Hungerländer, P., Rendl, A., & Truden, C. (2017). On the Slot Optimization Problem in On-Line Vehicle Routing. Transp. Res. Proc., 27, 492–
499. https://doi.org/10.1016/j.trpro.2017.12.046
Jacobsen-Grocott, J., Yi Mei, Gang Chen, & Zhang, M. (2017). Evolving heuristics for Dynamic Vehicle Routing with Time Windows using
genetic programming. 2017 IEEE Congress on Evolutionary Computation (CEC), 1948–1955. https://doi.org/10.1109/CEC.2017.7969539
Jia, Y.-H., Chen, W.-N., Gu, T., Zhang, H., Yuan, H., Lin, Y., Yu, W.-J., & Zhang, J. (2018). A Dynamic Logistic Dispatching System With Set-
Based Particle Swarm Optimization. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 48(9), 1607–1621.
https://doi.org/10.1109/TSMC.2017.2682264
Liao, T. Y., & Hu, T. Y. (2011). An object-oriented evaluation framework for dynamic vehicle routing problems under real-time information.
Expert Syst. Appl., 38(10), 12548–12558. https://doi.org/10.1016/j.eswa.2011.04.041
Lim, K. L., Whitehead, J., Jia, D., & Zheng, Z. (2021). State of data platforms for connected vehicles and infrastructures. Commun. Transp. Res.,
1, 100013.
Nazari, M., Oroojlooy, A., Snyder, L., & Takác, M. (2018). Reinforcement learning for solving the vehicle routing problem. Adv. Neural Inf.
Process Syst., 9839–9849. https://doi.org/https://dl.acm.org/doi/10.5555/3327546.3327651
Okhrin, I., & Richter, K. (2008). The Real-Time Vehicle Routing Problem. In Operations Research Proceedings 2007 (pp. 141–146). Springer
Berlin Heidelberg. https://doi.org/10.1007/978-3-540-77903-2_22
Toth, P., & Vigo, D. (2014). Vehicle Routing: Problems, Methods, and Applications. Society for Industrial and Applied Mathematics.
Xia, N., Cheng, L., Chen, S., Wei, X., Zong, W., & Li, M. (2018). Accessibility based on Gravity-Radiation model and Google Maps API: A case
study in Australia. J. Transp. Geogr., 72, 178–190. https://doi.org/10.1016/j.jtrangeo.2018.09.009
Yin, R. K. (2003). Case study research: Design and methods. Applied social research methods series (3rd ed.). SAGE Publications Inc.
Yu, G., & Yang, Y. (2019). Dynamic routing with real-time traffic information. Oper. Res., 19(4), 1033–1058. https://doi.org/10.1007/s12351-017-
0314-9
Zhang, H., Ge, H., Yang, J., & Tong, Y. (2021). Review of Vehicle Routing Problems: Models, Classification and Solving Algorithms. Arch.
Comput. Methods Eng. https://doi.org/10.1007/s11831-021-09574-x