Solving the Vehicle Routing Problem

In the Open-Source Software ‘ODL Studio’

Stančić, Nikola; Kovačević, Jovan; Cvijetinović, Željko;
Brodić, Nenad; and Mihajlović, Dragan


Abstract: Transport management is one of the done by companies with the goal to visit specified
most important logistical tasks, and designing the locations where some activity needs to be carried
optimal transport routes plays a major role in that out (e.g., delivery of goods). Routing and
process. Vehicle routing is primarily of interest to scheduling of vehicles have a significant share in
companies that organize the transportation of
goods to end users, intending to minimize the total the total operating costs of distribution and
cost of the routes. Being a well-known NP-hard transportation systems. To reduce such costs,
problem, the theoretical background of the vehicle operational research techniques are successfully
routing problem has been researched for many used in many situations. A route includes the
years and it has been the basis for the complete path traveled by a vehicle starting from
development of a large number of commercial the central depot, visiting the designated
software. However, affordable software solutions
are lacking. Difficult access to high-quality and locations and finally returning to the central
reliable spatial data related to the road network is depot. The task should be completed within a
also a limitation. This paper discusses the specified time and using the existing road
possibilities of applying open-source software for network. At the same time, all user requests must
solving the VRP. Open spatial data from a be met, and all imposed restrictions must be
collaborative project of the virtual community was respected.
used as the basis for modeling the road network
graph. The experiment was carried out on data on
vehicles and customers of a certain retail chain The methods for solving transport problems
and refers to the territory of Serbia. are based on some of the typical spatial analyses
and vehicle routing is primarily a spatial problem.
Index Terms: GIS, location-based services, open The importance of the Geographic Information
data, optimal routes, VRP System (GIS) as a sophisticated technology is
reflected in the collection, storage, management,
1. INTRODUCTION analysis and presentation of spatial data.

T HE Vehicle Routing Problem (VRP) is the

process of determining the set of optimal
routes in the road network along which the
Location-based computing (LBC) also plays an
important role in the optimization of delivery
vehicle routes. Its infrastructure consists of a
vehicles should move when serving a set of distributed mobile computing environment in
customers. It represents a generalization of the which each mobile device is aware of its location
travelling salesman problem (TSP). Vehicle and wirelessly connected, which enables
routing is present in various fields, most notably Location-Based Services (LBS) [8]. LBS are
in the optimization of the transport of people or, made up of a wide range of services that
more often, goods. This process is most often combine the device's current location (usually
obtained from global navigation satellite systems)
Manuscript received September 26, 2022. with context data (e.g., information about vehicle
N. Stančić is at the Department of Geodesy and behavior and driving parameters). The
Geoinformatics, Faculty of Civil Engineering, University of
Belgrade, Serbia combination of GIS and LBS technologies in
(e-mail: [email protected]) – contact person. solving the VRP is an established part of the
J. Kovačević is at the Department of Geodesy and procedure in which these two technologies are
Geoinformatics, Faculty of Civil Engineering, University of
Belgrade, Serbia (e-mail: [email protected]). used for data collection and as decision-making
Ž. Cvijetinović is at the Department of Geodesy and tools [14].
Geoinformatics, Faculty of Civil Engineering, University of
Belgrade, Serbia (e-mail: [email protected]).
N. Brodić is at the Department of Geodesy and Theoretical frameworks related to various
Geoinformatics, Faculty of Civil Engineering, University of optimization methods for solving logistics
Belgrade, Serbia problems have been thoroughly researched and
(e-mail: [email protected]).
D. Mihajlović is at the Department of Geodesy and
well-known for many years [1–2, 11, 13].
Geoinformatics, Faculty of Civil Engineering, University of However, despite this, in practice, many
Belgrade, Serbia (e-mail: [email protected]).

5
companies still apply experiential and manual o additional constraints.
methods at the operational level of optimizing the In order to determine the total costs of
transport of goods. There are several reasons for transportation along a certain route, it is
this situation, primarily the lack of affordable necessary to know the costs of travel between
software solutions. Therefore, the main idea of individual customers, as well as between
this paper is to explore the possibility of creating customers and depots. The existing road network
optimal transport routes using open-source tools is represented by a complete graph, where arcs
and open data, available to everyone. represent roads, and nodes represent
intersections, depots or customers. For each pair
The rest of the paper is organized as follows. of points and on the complete graph, an arc
In Section 2, the formulation of the Vehicle is associated with a weight that is directly
Routing Problem is given. Section 3 lists related to the distance and/or travel time of the
established approaches for solving VRP. Section vehicle. All weights are stored within the weight
4 describes the proposed open-source and open- matrix that can be either symmetric or
data approach. Finally, Section 5 summarizes the asymmetric. The weight of the entire route is then
results and suggests future research directions. calculated as the sum of the weights of all
segments that belong to the route.
2. PROBLEM DEFINITION
The vehicle routing problem was first The most common model of the VRP is
presented by George Dantzig and John Ramser characterized by capacity limitations and is called
in 1959 [6], calling it the truck dispatching the Capacitated Vehicle Routing Problem
problem. Current VRP models, however, are (CVRP). In it, each vehicle has a defined
immensely different from the original one, as they maximum load value (its carrying capacity), and
increasingly aim to incorporate real-life for each customer, information is kept on the
complexities, such as time-dependent travel number of goods to be delivered or collected
times (reflecting traffic congestion), time windows from him. The number of goods to be delivered to
for pickup and delivery, and input information customers in one route must not exceed the
(e.g., demand information) that changes maximum carrying capacity of the vehicle.
dynamically over time [3].
2.1. Variants of the VRP with capacity constraints
Solving the VRP is equal to the minimization of Depending on the additional restrictions that
total costs, that is, the goal is to find an optimal are introduced, there is a wide range of CVRP
set of feasible routes that require minimum costs variants. The basic CVRP adopts several
in terms of mileage, fuel consumption, required assumptions, such as the use of a homogeneous
time, number of used vehicles or some other vehicle fleet, single depot, one route per vehicle
criteria. Every vehicle (i.e., driver) is assigned a rule, etc. These limitations can be removed by
route and a schedule. In addition to achieving introducing additional restrictions.
cost minimization, all constraints and
requirements have to be satisfied: working hours, A simpler case is one in which there are no a
the number of mandatory stops, maximum travel priori set limits for the delivery time (such as the
time, special user requirements, etc. situation with, e.g., newspaper delivery). If,
however, the time of the customer visit is of
Input data required for solving the VRP primary importance and explicitly defined (as, for
includes: example, in the case that customers accept
o the starting and ending point(s) deliveries only within a certain time), then time
representing the depot(s); constraints can no longer be ignored and need to
o available fleet and vehicles be taken into account when routing and creating
characteristics; time schedules. Also, a priority restriction can be
o a set of nodes with known locations introduced that favors individual customers when
that represent the customers to be planning the order of serving customers. One
visited; type of priority restriction is the requirement that a
o existing road network; particular customer is served on a route that
o quantity of goods (there may be includes a particular subset of other customers.
several types of goods that need to be Another possibility is to require that a specific
collected or delivered); customer is served before (or after) a subset of
o the time window for each customer in customers. A limitation of this type represents the
which it is necessary to serve him; problem of pickup and delivery, where goods
o time required for unloading or loading collected from one customer must be delivered to
of goods; another customer by the same vehicle. It is a

6
common request to serve several groups of obtain an approximate solution that can be found
customers in one route, knowing which group of quickly enough and is sufficiently accurate. Due
customers should be served first. If both pickup to all of the above, a heuristic approach is much
and delivery are carried out on the route, and the more often used for practical solutions, which
restrictions are related to the process of loading provides an approximately optimal solution, but in
and unloading of goods and where it is real-time.
undesirable to reorganize the load on the route,
all deliveries must be made before pickups. [4] 3.1. Approaches to solving the VRP
The approaches to solving the vehicle routing
Four characteristic variants of the vehicle problem and the historical development of the
routing problem with capacity constraints, methods used can be traced through four
classified depending on the additional constraint decades of the twentieth century. In the early
that is introduced, are: 1960s, only small problems involving 10 to 20
o Distance-constrained Vehicle Routing customers could be solved. In the late 1960s, a
Problem – DVRP; breakthrough occurred with the introduction of
o Vehicle Routing Problem with Time 3-opt and 2-opt route enhancement mechanisms.
Windows – VRPTW; Algorithms developed until the beginning of the
o Vehicle Routing Problem with seventies could solve only some of the problems
Backhauls – VRPB; that contained from 30 to 100 customers. In the
o Vehicle Routing Problem with Pickup 1970s, the concept of two-pass algorithms was
and Delivery – VRPPD. introduced, and the speed of algorithm execution
itself became more significant. The size of the
Additionally, the vehicle routing problem can be problems that are solved increased significantly,
observed from two perspectives, as a static or and in that decade, from 1970 to 1980, some of
dynamic problem. In static models of the vehicle the problems with 25 to 30 customers are solved
routing problem, it is assumed that all parameters optimally. In the 1980s, the rapid development of
describing the problem are known in advance exact problem-solving based on linear
and decisions are made only once, at the programming began. Apart from the exact one,
beginning of the planning process. In contrast to other approaches to solving the vehicle routing
static, dynamic models of the vehicle routing problem based on experience are proposed and
problem, the input information is only partially developed. Some of the problems with 50
known at the time of vehicle route planning. New customers are optimally solved. A new concept of
information may arrive during the planning metaheuristics is being formed at the general
process or during its execution, and also the level of solving optimization-combinatorial
original information may be modified. Because of problems based on many methods newly
this, it is not possible to get a definitive route in introduced during the 1990s, which are
one attempt [5], making this type of VRP significantly different from each other. Vehicle
significantly more difficult to solve. routing problems that were optimally solved in the
nineties reached up to 100 customers.
3. EXISTING SOLUTIONS
Solving the VRP is considered to be one of the Therefore, contemporary approaches to
most challenging combinatorial optimization solving the vehicle routing problem can be
tasks. The problem is classified as an NP-hard divided into three main groups:
(nondeterministic polynomial time) problem [10], o exact approach;
which means that the computational effort and o heuristic approach;
solving time required to solve this problem grow o metaheuristic approach.
exponentially or even factorially with the size of
the problem. Exact algorithms are only efficient in Since the number of possible routes for the
cases where the number of customers is general case of vehicle routing grows very
relatively small. According to data in [9], the limit quickly, the practical application of the exact
is up to 50 to 100 orders, depending on the approach is very limited. The algorithms
applied variant and the required response time. belonging to this group are the Column
Real problems that a company may have usually Generation method, the "Branch and Cut"
have significantly more customers, so they algorithm and Lagrange relaxation. They are
cannot be solved in an acceptable time using an primarily used in hybrid algorithms or special
exact approach. The calculation time with this cases when the set of possible solutions is
approach significantly exceeds the time in which significantly narrowed.
it is necessary to decide on a new route plan. The heuristic approach is based on the use of
Instead, for such problems, it is desirable to experience, user intuition and own judgment.

7
Unlike exact methods, heuristic methods try to scripts that represent separate modules, each of
reduce the number of possible routes in the which implements one specific functionality. It is
problem-solving process. Heuristic algorithms are possible to run pre-made scripts or create new
often based on heuristic route construction, ones using intuitive wizards. What makes ODL
where constructing and improving routes is done Studio distinctive is that, unlike classic GIS
iteratively. During this process, the customer is software and similar software, the data used by
required to actively engage, guide and correct the this application is in XLS format, i.e., in a simple
algorithm based on their knowledge, thereby Microsoft Excel spreadsheet. This enables the
improving the final output of the algorithm. Some data to be edited both in ODL Studio and in MS
of the representatives are the Clark-Wright Excel without any problems. Due to the simplicity
Algorithm, Sweep Algorithm, Nearest Neighbor of this format, interoperability with other systems
Heuristic (NNH), Farthest Addition Heuristic is also ensured.
(FAH) and the Christofides-Mingozzi-Toth
Algorithm. For vehicle routing purposes, ODL Studio uses
the free GraphHopper routing library written in
The third and most modern of the three Java and the open-source jsprit toolkit, which is
approaches to solving NP-hard problems of specialized for solving traveling salesman and
combinatorial optimization is a metaheuristic, vehicle routing problems. A detailed description
which, as a term in optimization, represents a set of the features, content, and API of the Java
of concepts that define heuristic methods used in GraphHopper routing library is available at [7].
solving a large number of different problems.
Metaheuristics in practice is a set of algorithms 4.1. Experiment
that are used to solve many different optimization The territory for which the optimization tasks
problems where the algorithm itself changes very were solved is the territory of Serbia. The road
little depending on the problem being solved. A network of Serbia is taken from OpenStreetMap
metaheuristic approach to solving VRP is often data that is compatible with the GraphHopper
based on local search guided by processes taken library and the ODL Studio software. Data
from nature [4]. Typical algorithms that have this regarding available vehicles and locations to be
approach are the method of Simulated Annealing visited are taken from the fleet management
(SA), Genetic Algorithms (GA), Tabu Search (TS) system database of the retail chain. The
and Ant Colony Optimization (ACO), while the constraints that were taken into account when
methods that apply Neural Networks (NN) are creating the optimal solution are expressed
increasingly being developed. through the corresponding columns in the input
tables. There are limited payloads of delivery
In addition to the three basic groups of vehicles, in addition to time windows that are
algorithms mentioned above, so-called hybrid defined for both vehicles and customer locations.
methods have also been developed. They use a Apart from that, there are two types of customers:
combination of at least two methods from the one part of customers requests delivery of goods
group of exact, heuristic and metaheuristic to their locations, while the other part of
procedures to solve vehicle routing problems. customers requests that a certain amount of
The hybrid approach is still relatively young and goods be picked up from their locations.
in its infancy and, as stated in the paper [9], Therefore, the solving of the capacitated vehicle
based on the available literature little research routing problem with time windows and
has been carried out using them. simultaneous pickup and delivery
(CVRPTWSPD) is illustrated.
4. PROPOSED OPEN-SOURCE SOLUTION
Open Door Logistics Studio or just ODL Studio GraphHopper models the road network of the
[12] is a free and open-source desktop software given territory in the form of a graph whose arcs
designed for finding optimal solutions for represent traffic roads, and nodes represent
problems in the field of logistics. It was developed intersections, depots or customers. The graph
by the British company Open Door Logistics in obtained in this way is saved in several different
2014 and is available under the GNU Lesser files and then used to calculate distances and
General Public License. Tools within it are travel times between nodes. The basis of the
developed for analysis and manipulation of data graph of the road network is represented by the
on user locations, shaping and mapping of the most up-to-date original OSM data, which were
territory where deliveries are carried out and – downloaded from the web server of the German
the key point for this paper – optimization of company Geofabrik in PBF format.
vehicle routing and creation of the order for ODL Studio requires the creation of three
visiting locations. ODL Studio is organized by tables: Stops, VehicleTypes and Stop-order,

8
where the first two tables contain the input data. handed over to a customer or goods should be
These tables should be inside a single XLS file, taken from him (e.g., return of goods). It was
so that each one is a separate spreadsheet in it, assumed that approximately 10% of customers
identically named. will request a certain amount of goods to be
picked up from their locations, while the other
90% will request delivery of goods to their
locations. To simulate this real scenario, the
basic functions of MS Excel were used.

The limits of the time window, start-time

and end-time, are set to identical values for all
users. By default, the user cannot receive the
goods before 5:30 a.m. (value '05:30:00'), as
well as that the delivery of the goods must be
done by 8:30 p.m. (value '20:30:00'). The
duration of the handover is simulated similarly to
the type column. The assumption is that
Figure 1: Importing the data on the fleet and the customers
vehicles will not stay at all locations for the same
The Stops worksheet contains a list of user amount of time. Delivering goods to a small shop
locations to visit. In addition to basic customer will usually involve unloading only a few boxes,
data (primarily delivery addresses and location which will take significantly less time compared to
coordinates), this worksheet also includes time delivering a larger quantity of goods to a
limits set by the customer, the duration of the act supermarket. In the simulation, it is predicted that
of handing over the goods, and the number of the value of the service-duration attribute
goods to be delivered. The purpose of the for approximately 60% of locations will be 10
second worksheet, VehicleTypes, is to store minutes, for about 20% of locations 20 minutes,
general information about the fleet and its and for the remaining 20% of locations 40
vehicles. The coordinates of the initial and final minutes. The quantities of goods requested by
depot, the working hours of the delivery person, customers (column quantity) are assigned
the carrying capacity of the vehicle and various through a simulation at the SQL level.
parameters used to calculate the driving costs
are stored (fixed costs, driving costs per A similar approach applies to the columns of
kilometer traveled and per elapsed hour, and the vehicle fleet characteristics table. Data on
waiting costs per elapsed hour). The number of vehicle identifiers and names, storage locations,
vehicles available in a specific warehouse is also as well as time windows for vehicles are obtained
known. The third Stop-order depot has the role of from the corresponding columns. The amount of
a connection table, i.e., each of its rows will carry goods for delivery that can be placed in each
information about which customer is assigned to vehicle (column capacity) depends on the type
which vehicle. This table is initially empty and its of vehicle, that is, whether it is a van, a small
filling is done after successfully finding a set of truck or a tow truck. Available vehicles are
optimal routes. classified by capacity based on their garage
number.
To populate the Stops and VehicleTypes tables
with real data, SQL scripts on the Microsoft SQL Additional abilities of drivers or vehicles are not
Server database of the trade chain were written. considered, so the value of the skills column
For some columns of the first table, equivalent is equal to NULL. Based on the demo input data
information was found in the database: location that can be created within ODL Studio for several
ID and name, as well as longitude and latitude in countries, the values of the parameters used to
the WGS84 system. The addresses were calculate the route costs have been entered.
obtained by the geocoding process based on the Thus, it was assumed that the amount of driving
above-mentioned coordinates and the available costs per kilometer traveled is '0.001', and per
address system. Filling the other columns with
elapsed hour is '1'. If the driver is forced to
meaningful values had to be simulated in another
spend a certain amount of time waiting due to
way. Since the job-id and required-skills
time windows, the waiting costs will amount to
columns do not need to have values, they are set
'0.5' per past hour. The part of the costs that
to NULL. The value of the type column can be
are always present and defined by the fixed-
'D' (for delivery) or 'P' (for pickup). Each row
cost parameter is 100.
of the table should receive one of these two
The ODL Studio software module intended for
values, depending on whether goods should be
solving vehicle routing problems is called Vehicle

9
routing & scheduling. The process of finding the The map is interactive, meaning that it is
optimal solution consists of the following possible to pan and zoom the view. By
activities: positioning the mouse cursor on an arbitrary
o import of prepared input data; location of the customer, insight into the address
o initialization of the tool for optimally and time when that location will be visited is
solving the vehicle routing problem; obtained. By enlarging the view to the level
o analysis of resulting routes; where the roads can be seen, it is noticeable that
o analysis of Gantt charts; the projected routes faithfully follow the road
o (optional) manual changing of routes; network.
o reports creation;
o final analysis and export. The dynamic plan for the realization of these
routes can be viewed using a Gantt chart. The
The created graph of the road network of vehicle IDs are listed on the vertical axis, and the
Serbia and the XLS file with data on available time is shown on the horizontal axis. By
vehicles and existing orders for goods should be positioning the mouse cursor on an arbitrary
loaded into the software. The next step is segment of the Gantt chart, information is
creating a new script for the Vehicle routing & obtained about the state of the specific vehicle at
scheduling module with 1 quantity type. This a specific moment (whether it is serving the
configuration was chosen because the input data customer, standing and waiting, or traveling).
refers to one-dimensional quantities of goods. Vehicle waits are the consequences of defined
The creation of optimal routes is triggered by the time windows.
Optimise option. The optimization process can
take from a few seconds to a few minutes, mostly
depending on the amount of data, i.e., the
number of locations that vehicles need to visit. In
this particular case, the time for processing the
input data was just over one minute. An overview
of the designed routes is generated by the View
routes in the map command from the View
solution menu. It can be seen that
OpenStreetMap is used as a background, and
also that the route of each vehicle is colored in a
different color.
Figure 4: Gantt chart of the obtained solution

A good solution has a small number of inactive

periods. The efficiency of the solution is reflected
in the degree of time utilization, i.e., the fastest
performance of the task (shown in green in the
Gantt chart), with the minimization of travel time,
which directly increases costs (blue segments in
the Gantt chart).

Although the software certainly provided the

optimal solution for the given input data, it must
be emphasized that this solution is optimal only
Figure 2: Display of all routes drawn over OpenStreetMap
with respect to the defined constraints. In
practice, things need not be so strict. Sometimes
it is possible to save resources in total if a slight
break of the rules is allowed (a slight exceeding
of the defined capacity or a small deviation from
the given time window). It is up to the operator to
decide whether this is acceptable to him, taking
into account other factors and his intuition. Due to
the above reasons, it is possible to manually
change routes: by excluding a particular location
from the solution by moving it to the category of
non-aligned locations or by transferring it to
Figure 3: Display of a single route drawn over
another or changing the order of visiting locations
OpenStreetMap
within one route. For example, if the company's

10
business strategy allows, it may be decided that also possible that the location was deliberately
a better option is to postpone the delivery of skipped by the decision of the company's
goods to that location to another day. Also, the management when manually changing the
manual modification could provide a more routes.
efficient solution for routes that overlap.
Reports in PDF format are a convenient form
The View loads command is very useful. It of a visual and statistical overview of the
creates a graph that shows the change in the optimized goods delivery plan. The report
number of goods in each vehicle over time. Each contains the name and identifier of the vehicle,
route is colored in a different color, and there is the number of locations within the route, the
an option to turn routes on and off by checking length of the trip in kilometers, a graphic
the appropriate boxes. It shows how the amount representation of the route on the OSM map and
of cargo in vehicles fluctuates over time. This is a a list of locations with data on the name and
consequence of different types of services – in address of the location and the time when the
some locations the goods are unloaded from the location will be visited.
vehicle, while in some locations it is necessary to
pick them up from the customer.

Figure 5: Graph of changes in the number of goods in

vehicles over time for three arbitrarily selected routes

In addition to the visual overview of the routes,

ODL Studio also offers a high-quality insight into
the statistical parameters that describe the
obtained solution in detail. All details about
planned deliveries, routes and the complete
timetable are available within commands under
the View solution tab. The View solution-details
command gives the table in which various data
about the entire solution are summarized. The
View route-details option provides a table with Figure 6: Example layout of one page of a route report
similar columns, with the difference that the Statistical tables can be viewed in ODL Studio,
statistical parameters are listed for each route but they will not be placed in MS Excel tables
separately. If there are irregularities in a route in until they are exported. The commands intended
the form of exceeding the limit, that route will be for export are located under the Export solution
colored red. The View stop-details command tables tab. After tables are saved, they can be
opens a table with statistics related to the opened and viewed independently of ODL
planned locations. This table is the most detailed. Studio. The exported data is in XLS format, so
View unassigned stops is used to list those these tables can be passed to some other
locations that are excluded from the delivery plan system that knows how to read them.
for various reasons. One possibility is that the
software itself skipped a certain location because 5. CONCLUSION
it is in an area for which there is no pre-prepared
road network graph. It should be checked that This paper considers the possibility of using
there is no error in the coordinates of the open-source software on open data to solve the
location, which can lead to a situation where a vehicle routing problem. In general, by allowing
location is mistakenly located in the wrong the computer to create routes for the delivery of
country and is too far from other locations. It is goods to user locations, multiple benefits are

11
achieved. Therefore, there is a real interest in [4] Carić, T., “Unapređenje organizacije transporta
primjenom heurističkih metoda”, PhD dissertation,
companies to implement routing principles in their Faculty of Transport and Traffic Sciences, University of
systems. The performed experiment showed that Zagreb, 2004.
ODL Studio with the incorporated GraphHopper [5] Chen, WN., “Application of queuing theory to dynamic
vehicle routing problem”, Global Journal of Business
library is a fairly simple and powerful tool for
Research, Vol. 3, No. 2, pp. 85-91, 2009.
efficiently solving typical logistics tasks. [6] Dantzig, GB., Ramser, JH., “The Truck Dispatching
Moreover, it can be said that the more Problem”, Management Science, Vol. 6, No. 1, pp. 80-
demanding part of the task is to prepare all the 91, 1959.
[7] GraphHopper, Accessed: 19 September 2022, Available:
input data appropriately, that is, to satisfy the https://www.graphhopper.com
required form. The steps of processing and [8] Karimi, HA., Hammad, A., “Telegeoinformatics: Location-
finding solutions are simple and logically follow based Computing and Services”, CRC Press, Boca
Raton, Florida, U.S.A., 2004.
each other. The software is easy to use, which [9] Kumar, SN., Panneerselvam, R., “A Survey on the
can be attributed to the clarity of the interface, but Vehicle Routing Problem and Its Variants”, Intelligent
also to the interactive tutorials from the official Information Management, Vol. 4, No. 3, 2012.
[10] Lenstra, JK., Kan, AR., “Complexity of vehicle routing
website that gradually and intuitively guide the and scheduling problems”, Networks, Vol. 11, No. 2, pp.
user through the entire procedure. There are 221-227, 1981.
different types of analysis of the obtained [11] Narciso, M., Piera, MA., Guasch, A., “A Methodology for
Solving Logistic Optimization Problems through
solution, and the processing speed is quite Simulation”, SIMULATION: Transactions of The Society
satisfactory. for Modeling and Simulation International, Vol. 86, No. 5,
pp. 369-389, 2010.
[12] Open Door Logistics Studio, Accessed: 19 September
The combination of ODL Studio with
2022, Available: https://www.opendoorlogistics.com
OpenStreetMap data provides a free, efficient, [13] Petrović, G., Madić, M., Antucheviciene, J., “An
robust and affordable alternative to commercial approach for robust decision-making rule generation:
software solutions for the same purpose. This Solving transport and logistics decision-making
problems”, Expert Systems with Applications, Vol. 106,
confirmed the initial hypothesis that open-source pp. 263-276, 2018.
software and free spatial data on the road [14] Vojinović, M., Cvijetinović, Ž., Đokić, M., Obradović, A.,
network are a solid basis for a system intended to “Primena sistema za nadzor voznog parka u
kompanijama koje upravljaju otpadom – primer JKP
solve the problems of organizing the Mediana Niš”, INFOTECH 2012, Vrnjačka Banja, 29-31
transportation of goods to end users. May 2012.

Nikola Stančić is a PhD student and works as a teaching

The quality of OSM data can compete with the assistant at the Department of Geodesy and Geoinformatics,
quality of data offered by numerous commercial Faculty of Civil Engineering, University of Belgrade. He
services and web maps. Their great advantage, graduated as a Bachelor of Geodesy in 2015 and defended
his master’s thesis in 2017. His main research areas are GIS,
which is also used by ODL Studio, is that they spatial data management, remote sensing and location-based
are completely open, and available to everyone services.
and anyone can edit and use them for their own
Jovan Kovačević, PhD works as an assistant professor at
needs. As the level of detail, completeness and
the Department of Geodesy and Geoinformatics, Faculty of
up-to-dateness of OSM data can vary depending Civil Engineering, University of Belgrade. He graduated as a
on the location, it is of great importance to Bachelor of Geodesy in 2015, defended his master’s thesis in
improve the quality of OpenStreetMap data and, 2017 and received his doctoral degree in 2022. His main
research areas are remote sensing, photogrammetry, GIS
in general, to further improve and promote this and machine learning.
project whose goal is a costless map of the whole
world. Željko Cvijetinović, PhD works as an associate professor at
the Department of Geodesy and Geoinformatics, Faculty of
Civil Engineering, University of Belgrade. He graduated as a
ACKNOWLEDGEMENT Bachelor of Geodesy in 1989 and received his doctoral
degree in 2005. His main research areas are digital terrain
This research was partially supported by the modeling, GIS, photogrammetry and LiDAR.
Serbian Ministry of Education, Science and
Technological Development under grant no. Nenad Brodić is a PhD student at the Department of
Geodesy and Geoinformatics, Faculty of Civil Engineering,
200092. University of Belgrade. He graduated as a Bachelor of
Geodesy in 2010 and defended his master’s thesis in 2012.
REFERENCES He works as a teaching assistant at the Faculty of Civil
Engineering, University of Belgrade. His main research areas
[1] Baker, BM., Ayechew, MA., “A genetic algorithm for the are LiDAR, GIS, remote sensing and spatial data
vehicle routing problem”, Computers & Operations management.
Research, Vol. 30, No. 5, pp. 787-800, 2003.
[2] Bin, Y., Zhong-Zhen, Y., Baozhen, Y., “An improved ant Dragan Mihajlović, PhD works as an associate professor at
colony optimization for vehicle routing problem”, the Department of Geodesy and Geoinformatics, Faculty of
European Journal of Operational Research, Vol. 196, Civil Engineering, University of Belgrade. He graduated as a
No. 1, pp. 171-176, 2009. Bachelor of Geodesy in 1980 and received his doctoral
[3] Braekers, K., Ramaekers, K., Van Nieuwenhuyse, I., degree in 1991. His main research areas are
“The vehicle routing problem: State of the art photogrammetry, remote sensing, GIS and 3D modeling.
classification and review”, Computers & Industrial
Engineering, Vol. 99, pp. 300-313, 2016.

12