ARCHIVES OF TRANSPORT ISSN (print): 0866-9546

Volume 67, Issue 3, 2023 e-ISSN (online): 2300-8830

DOI: 10.5604/01.3001.0053.7463

RISK MANAGEMENT IN THE ALLOCATION OF

VEHICLES TO TASKS IN TRANSPORT
COMPANIES USING A HEURISTIC ALGORITHM

Mariusz IZDEBSKI
Faculty of Transport, Warsaw University of Technology, Warsaw, Poland

Abstract:
The work deals with the issue of assigning vehicles to tasks in transport companies, taking into account the minimization
of the risk of dangerous events on the route of vehicles performing the assigned transport tasks. The proposed risk man-
agement procedure based on a heuristic algorithm reduces the risk to a minimum. The ant algorithm reduces it in the event
of exceeding the limit, which differs from the classic methods of risk management, which are dedicated only to risk assess-
ment. A decision model has been developed for risk management. The decision model considers the limitations typical of
the classic model of assigning vehicles to tasks, e.g. window limits and additionally contains limitations on the acceptable
risk on the route of vehicles' travel. The criterion function minimizes the probability of an accident occurring along the
entire assignment route. The probability of the occurrence of dangerous events on the routes of vehicles was determined
based on known theoretical distributions. The random variable of the distributions was defined as the moment of the vehi-
cle's appearance at a given route point. Theoretical probability distributions were determined based on empirical data
using the STATISTICA 13 package. The decision model takes into account such constraints as the time of task completion
and limiting the acceptable risk. The criterion function minimizes the probability of dangerous events occurring in the
routes of vehicles. The ant algorithm has been validated on accurate input data. The proposed ant algorithm was 95%
effective in assessing the risk of adverse events in assigning vehicles to tasks. The algorithm was run 100 times. The des-
ignated routes were compared with the actual hours of the accident at the bottom of the measurement points. The graphical
interpretation of the results is shown in the PTV Visum software. Verification of the algorithm confirmed its effectiveness.
The work presents the process of building the algorithm along with its calibration.
Keywords: risk management, heuristic algorithm, ant algorithm, optimization

To cite this article:

Izdebski, M. (2023). Risk management in the allocation of vehicles to tasks in transport
companies using a heuristic algorithm. Archives of Transport, 67(3), 139-153. DOI:
[Link]

Contact:
1) [Link]@[Link] [[Link]

Article is available in open access and licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0)
140 Izdebski, M.,
Archives of Transport, 67(3), 139-153, 2023

1. Introduction transport, or even loss of health and life of the trav-

The concept of risk is interpreted differently depend- ellers. The work focuses on risk management in the
ing on the area under study. The risk of accidents is issue of assigning vehicles to tasks.
often examined in rail (Szaciłło et al., 2021), air The presented approach to risk management in
(Izdebski et al., 2023), road (Huang et al., 2021) or transport companies is based on an approach that
intermodal (Kukulski et al., 2023) transport. The pri- considers the use of a heuristic algorithm in mini-
mary stage of risk management in various transport mizing dangerous events during the performance of
and logistics systems is calculating the probability transportation tasks by transport companies. The ad-
of random events (Zabielska et al., 2023; Semenov vantage of these algorithms is the quick time of gen-
& Jacyna, 2022). In road transport (Batarliene, 2020; erating the result, which is essential in the case of
Holeczek, 2019), risk means a measure of threat or determining vehicle driving routes and their quick
danger that may result either from probable events update (Agrawal et al., 2022; Ji et al., 2019; Ongcu-
beyond the control of a road user or from the possi- naruk et al., 2021).
ble consequences of making specific decisions. Risk The risk management procedure in transport compa-
in transporting goods is a set of factors that can dam- nies, according to which the assignment of vehicles
age the transported shipment. The risk in the to tasks was determined, consists of the classic
transport of people in activities can lead to the loss stages of the risk management procedure and addi-
of health and even the life of travellers. tionally considers the aspect of assigning vehicles to
Risk in transport companies is a complex decision transport charges.
problem. This is dictated by many aspects that need The stage of assigning vehicles to tasks aims to re-
to be analyzed and factors that affect the proper so- duce the risk in the transport company to an accepta-
lution to the problem. In general, the purpose of risk ble level As a result of assigning vehicles to tasks in
management in transport companies is to determine transport companies, vehicle routes are created. The
routes for vehicles that are characterized by a mini- tool controlling the level of risk is an optimization
mum probability of occurrence of various types of algorithm that assesses the risk of dangerous events
dangerous situations, e.g. traffic accidents or ran- on the route of vehicles. The algorithm calculates
dom weather events. new driving routes and allocations if the limit is ex-
It is worth emphasizing that reducing the risk of dan- ceeded. At the risk analysis stage, potential danger-
gerous events in transport companies increases the ous points are indicated on the route of vehicles for
efficiency of transport processes and contributes to which the probability of a dangerous event and its
increasing the attractiveness of road transport com- consequences is determined. Estimated probabilities
pared to other modes of transportation, e.g. air or rail are determined when the vehicle appears at a given
transport. Dangerous situations in transport compa- point on the route. It is therefore necessary to desig-
nies appear when carrying out the ordered transport nate routes for assigning vehicles to tasks in such a
tasks. It is, therefore, advisable to plan the transpor- way as to reduce the probability of dangerous events
tation of goods or people in such a way as to avoid at these points to a minimum. The optimization al-
collisions with other road users. Minimizing the risk gorithm evaluates the generated allocation and cal-
of dangerous events and, thus, risk management culates the likelihood of accident risk and its conse-
should occur when planning vehicle routes by plan- quences. If the acceptable level of risk is exceeded,
ners or organizers providing transport services. it rejects the developed solution and repeats the stage
In the transport process, various decisions are made of selecting the assignment of tasks to vehicles. The
with varying degrees of risk of performing the com- ant algorithm was used to manage the selection of
missioned task. The critical decisions affecting the vehicles for tasks. The ant algorithm belongs to the
quality and efficiency of the transport service are the group of heuristic algorithms that are often used in
decisions on the choice of the vehicle route and the complex optimization problems (Giovanni et al.,
assignment of these vehicles to tasks. In both cases, 2002). A valuable advantage of selected heuristic al-
the consequence of making a wrong decision is fail- gorithms is their short calculation time.
ure to perform the order, destruction of the means of The paper assumes that the probability of an acci-
dent on a given section of the transport network is
determined based on theoretical distributions based
Izdebski, M., 141
Archives of Transport, 67(3), 139-153, 2023

on data measured on these sections of the road net- 2. The state of the art
work. The fit of the empirical distribution to the the- 2.1. The assignment problem in transportation
oretical one was carried out using the STATISTICA companies
13 program. The problem of assigning vehicles to tasks is simi-
The complexity of transport processes means that larly an optimization problem (Tian et al., 2022; Wei
risk management in transport companies should be et al., 2021; Munapo, 2020). The issue of assigning
supported by appropriate tools adapted to the nature vehicles to tasks belongs to a broad group of prob-
of these processes. Risk management is the manage- lems of allocating resources to tasks. This issue is
ment of the route of vehicles in such a way as to known in the literature and interpreted differently
avoid dangerous events with other road users or haz- depending on the problems under consideration (Yu
ardous situations. The analysis of current ap- et al., 2023; Dhouib, 2022). This issue involves as-
proaches to risk management methods and tools in signing available resources, e.g. vehicles and em-
allocating vehicles to tasks in transport companies ployees, to assigned tasks. The classic assignment
has shown that there are no models and algorithms assumes that each task, if possible, is assigned to
to solve the problem comprehensively, determining precisely one performer, and each performer com-
the risk of dangerous events along the entire pletes only one task. The measures of a correctly
transport route, not only in its specific place. It generated assignment usually determine the mini-
should be noted that none of the risk assessment and mum execution time for all tasks or the minimum
management tools allows for assigning vehicles to cost of task execution. The issue is often modified
tasks in such a way as to minimize this risk. From by introducing various combinations of the number
the point of view of research on risk management in of tasks assigned to resources, e.g. an equal number
transport companies, an essential aspect is not only of tasks and resources, more tasks than resources, or
the formulation of a model adequate to a given situ- fewer tasks than resources. In general, the assign-
ation but also the selection of an appropriate algo- ment problem is the association of resources with
rithm to solve it. tasks (Karsu et al., 2021). The consequence of this
The primary purpose of the work is to develop a tool association is a specific benefit that gets worse or
to support decision-making in the allocation of vehi- better depending on the combination chosen. In
cles to tasks in transport companies, considering transport issues, the issue of allocation becomes
minimizing the risk of dangerous events in routes more complex. There are constraints on driving
generated by the assignment. A new approach pre- time, work time and task completion time, making
sented in the work is to consider the risk aspect in assignments even more difficult. The single contrac-
assigning vehicles to tasks. tor constraint has been modified to allow multiple
The research presented in the work is contained in tasks to be performed by one vehicle. The number of
six chapters. The first chapter gives an introduction vehicles is only determined after assigning them to
to the undertaken research topic. The second chapter tasks. If the risk factor is taken into account in the
presents an overview of the literature on the issue of assignment of vehicles to tasks, the limit for exceed-
assigning vehicles to tasks and methods of risk man- ing the acceptable risk level should be additionally
agement in transport companies. The third chapter taken into account.
describes the risk management model in a transport The allocation problem also plays a fundamental
company and presents its formal record. The fourth role in determining transport routes (Lyu & Andrew,
chapter describes the optimization algorithm used to 2021; Fuentes et al., 2021). The allocation problem
reduce the risk of dangerous events on routes gener- also determines the minimum number of resources,
ated by assigning vehicles to tasks. Verification of e.g. vehicles, to complete all routes. An appropriate
the algorithm and its calibration are presented in the set of tasks, e.g., routes to be completed, is deter-
fifth chapter. The summary emphasizes the original- mined for each means of transport. The allocation
ity of the proposed research and its further direction. measure is the minimum length of the route of the
vehicles performing all the ordered transport tasks.
Hence, determining the minimum access routes to
individual tasks and loading routes is decisive in the
allocation problem. In resource allocation problems,
142 Izdebski, M.,
Archives of Transport, 67(3), 139-153, 2023

tasks take different forms, but the meaning of the and therefore are not considered for risk estimation,
problem is always the same. Tasks should be com- e.g. lightning strikes. Certain dangerous circum-
bined with available resources so that the benefit stances, e.g., vehicle breakdowns or driver errors,
from this combination is as good as possible. An es- are among the most common events contributing to
sential assignment in transport issues is the problem road accidents. The frequency of these events can be
of determining work schedules. Scheduling can be reduced through thorough technical inspections of
defined as the time allocation of available resources vehicles, or ensuring adequate rest for drivers, e.g.
to tasks. The assignment characterized in scheduling introducing a two-person crew for a given transport
differs from the classic assignment model, where task. An essential mechanism for reducing the num-
one agent (device, employee) performs precisely ber of accidents is the determination of such driving
one task and one task is performed by only one routes characterized by a minimum probability of
agent. The sense of scheduling is to determine the dangerous situations independent of the person per-
order of tasks and assign them to the performers of forming transport tasks, e.g., threats from other road
these tasks (employees, devices) so that these per- users.
formers complete these tasks in the minimum time. In road transport, the main threats are traffic acci-
One contractor can perform tasks sequentially in a dents caused by various factors depending or not on
given period. The problem of assigning vehicles to the driver performing the transport task (Ebrahim et
tasks is a complex decision problem belonging to al., 2021; Mahdi et al., 2020). The effect of an acci-
NP-hard problems. To solve it, it is required to use dent has a different weight depending on the size of
artificial intelligence algorithms, e.g. genetic algo- losses, damage, or the number of injured or killed
rithms (Jia et al., 2018). (Mujalli et al., 2023; Mokhtarimousavi et al., 2020).
In most cases, determining the risk in transport com-
2.2. Risk management in transport companies panies is based on analysing historical data on acci-
In transport companies, the risk is very often exam- dents, such as their frequency, consequences, and
ined in the context of transporting dangerous goods identifying factors contributing to their occurrence
and minimizing the number of accidents on the route (Hossaina et al., 2019).
of vehicles (Mohri et al., 2022). The transport of When making decisions under risk conditions in the
hazardous goods is a specific type that requires the transport of cargo or passengers, confident choices
development of such technology and transport or- are made, leading to various consequences, and it is
ganization that will minimize the probability of an essential to be able to assess the likelihood of these
accident risk on a given transport route (Hosseini & consequences. Therefore, determining the probabil-
Verma 2021; Timajchi et al., 2019). The conse- ity of a dangerous event and its implications is criti-
quences of accidents in transporting dangerous cal in risk estimation. The measure of risk is, thus, a
goods may lead to loss of health or life of people in combination of the size of the possible loss and the
the danger zone, the need to immediately evacuate probability of incurring this loss (Stojanovic et al.,
people from hazardous areas, contamination of air, 2023).
water and soil, degradation of the natural environ- The risk in road transport depends on the choice of
ment, and severe material losses. the vehicle route (Haixing & Qiangian, 2022; For-
Research (Huang et al., 2021) shows that the leading nalchyk et al., 2021), so it is necessary to designate
cause of accidents in transporting dangerous goods such a route to minimize dangerous situations along
should be sought in the human factor. Accidents are the entire vehicle route. The routing issue is an opti-
caused by bravado, alcohol, stress, and weather con- mization issue, so it is advisable to use heuristic al-
ditions, but also by overloading the body in terms of gorithms.
health, mental and physical. Transporting dangerous The literature analysis confirmed that the problem of
goods is associated with the possibility of accidents allocation in transport companies is presented in a
causing fires, explosions or toxic environmental classical approach without considering the minimi-
contamination. To avoid such serious consequences, zation of the risk of accidents in routes generated by
mechanisms should be introduced to reduce the risk the realized allocation. The issue of assigning vehi-
of accidents in road transport. Certain random events
in road transport occur with a very low probability
Izdebski, M., 143
Archives of Transport, 67(3), 139-153, 2023

cles to tasks in transport companies is a complex de- the route at such a moment that the probability of
cision-making issue and requires the use of heuristic a dangerous situation and its consequences is min-
algorithms. imal.
- The vehicle's driving route is interpreted as a route
3. Risk management model in assigning vehi- to the first task, an optional return to the base, or
cles to tasks the execution of subsequent tasks and return to the
3.1. The assumption of the model base.
In the developed decision-making model, risk man- - Vehicles can be assigned to multiple tasks.
agement in assigning vehicles to tasks comes down - Tasks must be completed within one working day.
to assigning vehicles to transport tasks so that a min-
imum probability of dangerous situations on this 3.2. The data input and the decision variables
route characterizes the route generated due to this To develop a risk management model in the al-
assignment. The developed risk management model location of vehicles to tasks, the input data were de-
can transport cargo and passengers, depending on fined:
the defined transport task. A transport task in cargo - WB - A set of transportation bases.
transportation is defined as picking up a load from - Z - A set of loading points.
the loading point and transporting it to the unloading - WW - A set of unloading points.
point. In public transport, a transport task can be de-
- T - A set of time intervals.
fined as a communication line to which vehicles are
- Q - The size of the load collected at the sender.
assigned to serve it. The assignment in the model is
interpreted as a decision to assign a vehicle leaving - T1 - Loading time.
the base to the first transport task or to assign a ve- - T2 - Unloading time.
hicle completing the current task to the next one. A - T3 - Waiting time for loading.
random variable was introduced into the model to - T4 - Waiting time for unloading.
determine the theoretical distribution of the proba- - TPP1 - Travel time between drop-off points and
bility of a dangerous situation occurring on assign- pick-up points.
ment routes or task routes. The risk of a hazardous - TPP2 - Travel time between unloading points and
condition on the access routes to the tasks and the shipping points.
task routes depended on the moment of the vehicle's - TPP3 - Travel time between the base and the
appearance on these routes. Therefore, the task of points of departure.
the risk management model in assigning vehicles to - TPP4 - Travel time between the landing point and
tasks is to assign cars to tasks at such moments when the base.
the risk is minimal. The developed risk management - PP1 - The probability of a dangerous event occur-
model is an optimization model with a designated ring between the points of origin and the points of
objective function minimizing the risk of dangerous receipt.
situations along the entire vehicle route. In addition, - PP2 - Probability of a dangerous event occurring
the model determines the minimum number of vehi- between unloading points and shipping points.
cles of a particular type that should be used to com- - PP3 - The probability of a dangerous event occur-
plete the assigned tasks. To construct the risk man- ring between the base and the points of departure.
agement model, the following assumptions were - PP4 - The probability of a dangerous event occur-
made: ring between the unloading point and the base.
- Vehicle routes between the base and tasks and be- - PPS1 - Probability of the occurrence of the result
tween tasks are known. of a dangerous event between the points of origin
- Routes of transport tasks are known. and the points of receipt.
- The number of vehicles to be assigned is known. - PPS2 - Probability of the occurrence of the result
- The random variable determining the probability of a dangerous event between the unloading points
distribution of a dangerous event has an interpre- and the points of dispatch.
tation of the moment (time) of the occurrence of - PPS3 - Probability of the result of a dangerous
the event on a given section of the route. There- event between the base and the points of departure.
fore, the vehicle must appear on a given section of
144 Izdebski, M.,
Archives of Transport, 67(3), 139-153, 2023

- PPS4 - Probability of the result of a hazardous ∀(w,w') ∈ 𝑳𝑩𝒁, ∀𝑡 ∈ 𝑻, ∀st ∈ 𝐒𝐓, ∀𝑧𝑎𝑑 ∈ 𝒁𝑨𝑫
event occurring between the unloading point and 𝑥𝑝1((w,w'), 𝑧𝑎𝑑,st,t) ∙ 𝑣(st) ≤ 𝑞(𝑧, 𝑧𝑎𝑑) (1)
the base. 𝑥𝑝2((w,w'), 𝑧𝑎𝑑,st,t) ∙ 𝑣(st) ≤ 𝑞(𝑧, 𝑧𝑎𝑑) (2)
- ST - A set of means of transport. 𝑥𝑝1((w,w'), 𝑧𝑎𝑑,st,t) ∙ ld(st) ≤ 𝑞(𝑧, 𝑧𝑎𝑑) (3)
- V - Vehicle capacity.
𝑥𝑝2((w,w'), 𝑧𝑎𝑑,st,t) ∙ ld(st) ≤ 𝑞(𝑧, 𝑧𝑎𝑑) (4)
- LD - Vehicle payload.
- TBP - Vehicle parking time.
- Limitation for the duration of tasks for a single
- TDP - The time allowed for the task to be com- vehicle:
pleted by the vehicle.
- A1 - The lower bound of the time window for the ∀(w,w') ∈ 𝑳𝑩𝒁, ∀𝑡 ∈ 𝑻, ∀st ∈ 𝑺𝑻, ∀(w,w') ∈
base. 𝑳𝑾𝑩
- B2 - The upper limit of the time window for the 𝑥𝑝1((w,w'), 𝑧𝑎𝑑,st,t)
base.
∙ [𝑡𝑝𝑝1(w,w',t) + 𝑡1(𝑧, 𝑧𝑎𝑑) + 𝑡3(z,t,zad)]
To formally record the risk management model in
the assignment of vehicles to tasks, it is necessary to + ∑ ∑ 𝑥𝑝4((w,w'), 𝑧𝑎𝑑,st,t)
define the decision variables of the model. The first (w,w')∈𝑳𝒁𝑾 𝑧𝑎𝑑∈𝒁𝑨𝑫 (5)
type of variables are binary variables describing ∙ [𝑡𝑝𝑝4(w,w',t) + 𝑡2(𝑧, 𝑧𝑎𝑑) + 𝑡4(z,t,zad)]
connections between network objects carried out by + ∑ ∑ 𝑥𝑝2((w,w'), 𝑧𝑎𝑑,st,t)
a given vehicle in a specific time interval and as- (w,w')∈𝑳𝑾𝒁 𝑧𝑎𝑑∈𝒁𝑨𝑫
signed to a given task. The second type of variable ∙ [𝑡𝑝𝑝2(w,w',t) + 𝑡1(𝑧, 𝑧𝑎𝑑) + 𝑡3(z,t,zad)]
determines the moment of departure of vehicles
+ 𝑥𝑝3((w,w'),st,t) + tbp(st) ≤ tdp(st)
from the base for the ordered transport tasks. The
purpose of introducing such types of decision varia-
bles is the same as in the risk management model in - Restriction on meeting the condition of acceptable
cargo transport. The driving time of vehicles should risk on the road:
be regulated so that the vehicles appear on a given
section at the time, generating the minimum proba- ∀(w,w') ∈ 𝑳𝑩𝒁, ∀𝑡 ∈ 𝑻, ∀st ∈ 𝑺𝑻, ∀(w,w') ∈
bility of a dangerous event. 𝑳𝑾𝑩
To develop a risk management model in the alloca-
tion of vehicles to tasks, the decision variables were [1 − 𝑥𝑝1((w,w'), 𝑧𝑎𝑑,st,t)
defined:
- XP1 - Connection between drop-off points and
∙ [1 − 𝑝𝑝E1(w,w') ∙ 𝑝𝑝𝑠E1(w,w')]
pick-up points.
- XP2 - Connection between unloading points and ∙ ∏ ∏ 𝑥𝑝4((w,w'), 𝑧𝑎𝑑,st,t)
shipping points. (w,w')∈𝑳𝒁𝑾 𝑧𝑎𝑑∈𝒁𝑨𝑫 (6)
- XP3 - Connection between the base and the points ∙ [1 − 𝑝𝑝E4(w,w') ∙ 𝑝𝑝𝑠E4(w,w')]
of origin. ∙ ∏ ∏ 𝑥𝑝2((w,w'), 𝑧𝑎𝑑,st,t) ∙ [1
- XP4 - Connection between the landing point and (w,w')∈𝑳𝑾𝒁 𝑧𝑎𝑑∈𝒁𝑨𝑫
the base. − 𝑝𝑝E2(w,w') ∙ 𝑝𝑝𝑠E2(w,w')] ∙ 𝑥𝑝3((w,w'),st,t)
- HP - The moment of departure of the vehicles
from the transport base.
∙ [1 − 𝑝𝑝E3(w,w') ∙ 𝑝𝑝𝑠E3(w,w')]] ≤ Rdo
3.3. Limits and the risk assessment function
Assignment constraints take the form:
- Limitation of the allocation resulting from the ca-
pacity and load capacity of the vehicles:
Izdebski, M., 145
Archives of Transport, 67(3), 139-153, 2023

- Limitation on fulfilment of time windows: 4. Ant algorithm in risk management in assign-

ing vehicles to tasks
∀(w,w') ∈ 𝑳𝑩𝒁, ∀𝑡 ∈ 𝑻, ∀st ∈ 𝑺𝑻, ∀(w,w') ∈ The concept of an artificial ant was introduced to de-
𝑳𝑾𝑩, w ∈ 𝑾𝑩 fine the ant algorithm. Each ant from the population
hp(st,w,zad) + 𝑥𝑝1((w,w'), 𝑧𝑎𝑑,st,t) creates its route with the interpretation of the assign-
∙ [𝑡𝑝𝑝1(w,w',t) + 𝑡1(𝑧, 𝑧𝑎𝑑) + 𝑡3(z,t,zad)] ment of vehicles to tasks. Waypoints are located in
three layers. The route takes a given vehicle's depar-
+ ∑ ∑ 𝑥𝑝4((w,w'), 𝑧𝑎𝑑,st,t) ture from the base at a specific time to the tasks and
(w,w')∈𝑳𝒁𝑾 𝑧𝑎𝑑∈𝒁𝑨𝑫 then returns to the base. Route elements are located
∙ [𝑡𝑝𝑝4(w,w',t) + 𝑡2(𝑧, 𝑧𝑎𝑑) + 𝑡4(z,t,zad)] (7)
in layers:
+ ∑ ∑ 𝑥𝑝2((w,w'), 𝑧𝑎𝑑,st,t) - Layer I: vehicles to be assigned.
(w,w')∈𝑳𝑾𝒁 𝑧𝑎𝑑∈𝒁𝑨𝑫 - Layer II: points on the interpretation of the vehi-
∙ [𝑡𝑝𝑝2(w,w',t) + 𝑡1(𝑧, 𝑧𝑎𝑑) + 𝑡3(z,t,zad)] cle's departure time from the transport base.
+ 𝑥𝑝3((w,w'),st,t) + tbp(st) ≥ 𝑎1(𝑤𝑏) - Layer III: tasks to be assigned.
The number of points in layer I depends on the ve-
hicles allocated. The number of elements in layer II
hp(st,w,zad) + 𝑥𝑝1((w,w'), 𝑧𝑎𝑑,st,t) is determined the same way as in the risk manage-
∙ [𝑡𝑝𝑝1(w,w',t) + 𝑡1(𝑧, 𝑧𝑎𝑑) + 𝑡3(z,t,zad)] ment model in cargo transport, i.e. on the number of
+ ∑ ∑ 𝑥𝑝4((w,w'), 𝑧𝑎𝑑,st,t) potential departure hours in a given period. The
(w,w')∈𝑳𝒁𝑾 𝑧𝑎𝑑∈𝒁𝑨𝑫 number of points in layer III depends on the number
∙ [𝑡𝑝𝑝4(w,w',t) + 𝑡2(𝑧, 𝑧𝑎𝑑) + 𝑡4(z,t,zad)] of tasks to be completed.
(8) The starting point of the ant route is the vehicle base.
+ ∑ ∑ 𝑥𝑝2((w,w'), 𝑧𝑎𝑑,st,t)
The other route of the ant, and thus the selection of
(w,w')∈𝑳𝑾𝒁 𝑧𝑎𝑑∈𝒁𝑨𝑫
subsequent route points, occurs with a certain prob-
∙ [𝑡𝑝𝑝2(w,w',t) + 𝑡1(𝑧, 𝑧𝑎𝑑) + 𝑡3(z,t,zad)] ability:
+ 𝑥𝑝3((w,w'),st,t) + tbp(st) ≤ 𝑏1(𝑤𝑏)
  τ (t)  α   η (t)  β , z   mr
  yz   yz 
PR mr yz (t) = 
 mr  τ yl (t)  ηyl (t)
α β
The criterion function minimizes the probability of (10)
dangerous events in the routes of vehicles:  lΩ  , z   mr
0
∀(w,w') ∈ 𝑳𝑩𝒁, ∀𝑡 ∈ 𝑻, ∀st ∈ 𝑺𝑻, ∀(w,w') ∈ where:
𝑳𝑾𝑩  yz (t ) – the intensity of the pheromone trace be-
𝐹(𝐗𝟏, 𝐗𝟐, 𝐗𝟑, 𝐗𝟒) tween the y-th point of the ant track and
the z-th point in the t-iteration of the algo-
= [1 − 𝑥𝑝1((w,w'), 𝑧𝑎𝑑,st,t) rithm,
η yz (t) – heuristic information:
∙ [1 − 𝑝𝑝E1(w,w') ∙ 𝑝𝑝𝑠E1(w,w')] 1
𝜂yz (𝑡) = (11)
𝑝(y,z) ⋅ 𝑝𝑠(y,z)
∙ ∏ ∏ 𝑥𝑝4((w,w'), 𝑧𝑎𝑑,st,t) where:
(w,w')∈𝑳𝒁𝑾 𝑧𝑎𝑑∈𝒁𝑨𝑫 (9) p(y,z) – the probability of an event on network sec-
∙ [1 − 𝑝𝑝E4(w,w') ∙ 𝑝𝑝𝑠E4(w,w')] tions,
∙ ∏ ∏ 𝑥𝑝2((w,w'), 𝑧𝑎𝑑,st,t) ∙ [1 ps(y,z) – the probability of the event's effect on net-
(w,w')∈𝑳𝑾𝒁 𝑧𝑎𝑑∈𝒁𝑨𝑫 work sections,
− 𝑝𝑝E2(w,w') ∙ 𝑝𝑝𝑠E2(w,w')] ∙ 𝑥𝑝3((w,w'),st,t)  ,  – the influence of pheromones and heuristic
data on the behavior of ants,
∙ [1 − 𝑝𝑝E3(w,w') ∙ 𝑝𝑝𝑠E3(w,w')]] ⟶ 𝑚𝑖𝑛  mr – the set of all point elements of the transport
network, l - potential ant route points taken
into account when selecting the next ant route
point.
146 Izdebski, M.,
Archives of Transport, 67(3), 139-153, 2023

In subsequent iterations, the pheromone trace is cal- runs until the stop condition is reached. The stop
culated according to the formula: condition is a fixed number of iterations. The num-
MR
ber of ants in the population creating individual ve-
 yz (t + 1) = (1 −  ) yz (t) +   yz mr (t) (12) hicle routes (solutions) and the number of iterations
mr =1 is set at the beginning of the algorithm implementa-
where: tion.
mr – another ant in the anthill mr ∈ 𝑴𝑹, The main steps of the ant algorithm can be presented
𝜌 – pheromone volatilization factor (0 < 𝜌 ≤ 1), as:
𝜏𝑦𝑧 (𝑡 + 1) – pheromone amplification, for the first - Step 1 - The first ant in the population starts mak-
iteration takes the value on each connec- ing a route.
tion equal 𝜏0 . - Step 2 - Selection of subsequent route points ac-
cording to the defined probability (10) until the ant
The first component of the formula (12) determines reaches the end point of the route (fulfillment of
the pheromone volatilization rate, while the second all tasks).
determines the pheromone gain and takes the value: - Step 3 - Repeat steps 1-2 for the next ant in the
1 population.
− 𝐾1mr (𝑡) − 𝐾2mr (𝑡) − - Step 4 - Pheromone update (12).
𝑃mr (𝑡)
𝑚𝑟
𝛥𝜏𝑦𝑧 (𝑡) = 𝐾3mr (𝑡) − 𝐾4mr (𝑡) - Step 5 - Repeating steps 1-4 until the algorithm
reaches the stopping condition. Selection of the
{ (13) ant route with the highest pheromone intensity
0
among all routes generated in the population. This
when segment (y,z) is used by the ant route is the final solution and determines the allo-
otherwise 0; cation of vehicles to tasks.
where:
Pmr(t) – the probability of the occurrence of a dan- 5. An example of using the ant algorithm for
gerous event and its effect in the entire route risk management in assigning vehicles to
created by the mr-th ant in the t-th iteration tasks
according to the formula (9); 5.1. The input data
K1mr(t) – penalty for exceeding time window limits To verify the optimization algorithms in the risk
(in model (7)-(8)) in the route created by the management model in the assignment of vehicles to
m-th ant in the t-th iteration of the algorithm, tasks, the transport network presented in Fig. 1 was
it is assumed that this penalty is half of the defined, in which the transport base and tasks were
pheromone accumulated on the route; determined. The tasks are carried out in the complete
K2mr(t) – penalty for exceeding the task completion truckload system and are characterized as picking up
time (5) in the route created by the m-th ant the load from the loading place and transporting it to
in the t-th iteration of the algorithm, it is as- the unloading point. Dangerous points of the route
sumed that this penalty is half of the phero- were determined in four places, and the time of ac-
mone accumulated on the route; cidents, collisions and road incidents was measured
K3mr(t) – penalty for exceeding the limit on vehicle in these places. For the remaining sections of the
capacity and load capacity (in model (1)-(4)) route, it was assumed that these sections were acci-
in the route created by the m-th ant in the t- dent-free. It was also believed that the capacity and
th iteration of the algorithm, it is assumed payload of a single vehicle are adapted to each task.
that this penalty is half of the pheromone ac- Tasks must be completed between 6:00 a.m. and
cumulated on the route; 5:00 p.m. The acceptable probability of a road acci-
K4mr(t) – penalty for exceeding the acceptable risk dent risk is 0.55. One time interval and one task to
limit (in model (6)) it is assumed that this be performed have been defined in the transport.
penalty is half of the pheromone accumu- Waiting times for loading have been omitted. Load-
lated on the route. ing and unloading time was set at 15 minutes. It was
The ant algorithm is an iterative algorithm, in the assumed that the execution time for all tasks is 2
next iteration its solution is improved. The algorithm hours.
Izdebski, M., 147
Archives of Transport, 67(3), 139-153, 2023

Fig. 1. Transport network with tasks

5.2. Theoretical distributions of the probability Table 1. Measuring points [hour expressed in
of accidents and their consequences in the minutes]
assignment of vehicles to tasks Point 1 Point 2 Point 4
Point 3 Type
Theoretical accident probability distributions were Type Type
715 C
Type
determined based on measurement data. The hours 712 C 422 C 710 C
745 C
of accidents at the measurement points are presented 743 S 621 S 740 C
728 C
720 C 472 C 721 C
in Table 1 (S - serious accident, C - collision). The 628 C
620 C 474 C 622 S
time of the accident is presented in minute format, 655 C 555 C
651 S
654 C
measuring period from 6:00 to 17:00. Theoretical 781 C
783 C 713 C 781 C
distributions of the accident random variable are 771 C
755 S 585 C 776 C
817 C
shown in Fig. 2. The linear scatterplot confirms the 812 C 661 S 813 C
833 S
normal distribution of the examined variable. The 835 C 674 C
872 C
834 C
chi-square test and the Kolmogorov-Smirnov (K-S) 642 C 782 C 1023 C
931 C
test were used to determine the distributions. 936 C 730 C 935 C
967 C
533 C 744 S 961 C
The null hypothesis stating the adopted type of dis- 871 C
878 C 722 C 872 S
tribution is rejected when the calculated value of the 972 C 811 S
757 C
754 C
statistics falls within the critical area determined by 793 C
754 S 857 C 788 C
the assumed significance level α=0.05. The values 843 C
793 C 642 C 849 S
of the Chi-square and Kolmogorov-Smirnov tests 977 C
844 S 913 C 970 C
865 C
and the parameters of the examined distributions are 872 C 681 C
662 S
859 C
presented in Table 2. 861 S 934 S 660 C
913 C
Tests of compliance for theoretical distributions of 664 C 942 S 916 C
844 S
the random variable of accidents classified as a col- 910 S 823 C 840 C
871 C
843 C 876 C 870 C
lision at individual measurement points are pre- 1010 C 1062 C
1061 S
550 S
sented in Table 3.
148 Izdebski, M.,
Archives of Transport, 67(3), 139-153, 2023

a) b)

c) d)
Fig. 2. Fitting the theoretical distribution of the random variable of the moment of road accidents at the
measurement points a) point 1 b) point 2 c) point 3 d) point 4

Table 2. Compliance testing and distribution parameters for road accidents of any type
Measurement Chi-square Distribution
points Statistics test (p) Statistics K-S test parameters Distribution
1 7,280 0,94 0,15 - μ=737; s2=34281 normal
2 9,898 0,77 0,05 - μ=723; s2=26941 normal
3 5,116 0,77 0,14 - μ=812; s2=22130 normal
4 1,620 0,89 0,11 - μ=805=; s2=14161 normal

Table 3. Compliance testing and distribution parameters for collision-type road accidents
Measurement Chi-square Distribution
points Statistics test (p) Statistics K-S test parameters Distribution
1 5,641 0,91 0,11 - μ=729; s2=33914 normal
2 4,983 0,91 0,12 - μ=723; s2=24128 normal
3 5,33 0,93 0,06 - μ=811; s2=22410 normal
4 2,67 0,90 0,04 - μ=817; s2=9921 normal
Izdebski, M., 149
Archives of Transport, 67(3), 139-153, 2023

5.3. Results of calibration and verification of the the ant assigned three vehicles to the assigned tasks.
ant algorithm in the assignment of vehicles A graphical presentation of the routes of the three
to tasks vehicles is shown in the Fig. 3 (green line – vehicle
The tested combinations of ant algorithm settings 1, blue line – vehicle 3, red line – vehicle 3). The
are shown in Table 4. Table 5 shows the aggregate departure time of the first vehicle was set at 8:15, the
results of the algorithm. Experimentally, the number second and third vehicles at 8:30. Based on Table 6,
of iterations of the algorithm was set at 200 repeti- it can be concluded that the efficiency of the ant al-
tions, and the population size at 100 ants. gorithm is 93%. The algorithm verification process
Table 5 shows that the minimum collision probabil- compares N algorithm solutions with actual data
ity value was 0.44 for test no. 24 with the algorithm specifying the times of dangerous events on route
parameters α=1, β=1, ρ=0.8. The route generated by sections.

Table 4. Test settings of the ant algorithm parameters

Test 𝜶 𝜷 𝝆 Test 𝜶 𝜷 𝝆 Test 𝜶 𝜷 𝝆
1 1 0,5 0,2 21 1 1 0,2 41 1 5 0,2
2 1 0,5 0,4 22 1 1 0,4 42 1 5 0,4
3 1 0,5 0,6 23 1 1 0,6 43 1 5 0,6
4 1 0,5 0,8 24 1 1 0,8 44 1 5 0,8
5 3 0,5 0,2 25 3 1 0,2 45 3 5 0,2
6 3 0,5 0,4 26 3 1 0,4 46 3 5 0,4
7 3 0,5 0,6 27 3 1 0,6 47 3 5 0,6
8 3 0,5 0,8 28 3 1 0,8 48 3 5 0,8
9 5 0,5 0,2 29 5 1 0,2 49 5 5 0,2
10 5 0,5 0,4 30 5 1 0,4 50 5 5 0,4
11 5 0,5 0,6 31 5 1 0,6 51 5 5 0,6
12 5 0,5 0,8 32 5 1 0,8 52 5 5 0,8
13 10 0,5 0,2 33 10 1 0,2 53 10 5 0,2
14 10 0,5 0,4 34 10 1 0,4 54 10 5 0,4
15 10 0,5 0,6 35 10 1 0,6 55 10 5 0,6
16 10 0,5 0,8 36 10 1 0,8 56 10 5 0,8
17 20 0,5 0,2 37 20 1 0,2 57 20 5 0,2
18 20 0,5 0,4 38 20 1 0,4 58 20 5 0,4
19 20 0,5 0,6 39 20 1 0,6 59 20 5 0,6
20 20 0,5 0,8 40 20 1 0,8 60 20 5 0,8

Table 5. Results of tests (P – probability of a collision in the route, T – running time [minutes:seconds])
Test P T Test P T Test P T Test P T Test P T
1 0,50 5:26 13 0,71 5:35 25 0,67 4:12 37 0,80 4:11 49 0,49 4:31
2 0,55 5:13 14 0,72 4:12 26 0,59 4:23 38 0,78 5:12 50 0,47 4:22
3 0,57 4:21 15 0,71 4:33 27 0,65 4:23 39 0,79 4:21 51 0,48 4:53
4 0,61 4:22 16 0,72 4:45 28 0,68 3:31 40 0,84 4:21 52 0,53 4:11
5 0,60 4:03 17 0,71 4:23 29 0,51 4:44 41 0,55 4:35 53 0,61 4:21
6 0,61 4:32 18 0,78 4:11 30 0,52 4:25 42 0,47 4:34 54 0,77 4:22
7 0,66 4:43 19 0,75 4:34 31 0,64 4:21 43 0,47 4:12 55 0,77 4:22
8 0,61 4:12 20 0,48 4:21 32 0,71 4:21 44 0,47 4:32 56 0,77 4:22
9 0,61 4:31 21 0,47 4:42 33 0,85 4:22 45 0,49 4:32 57 0,81 4:22
10 0,62 4:32 22 0,46 4:32 34 0,71 4:27 46 0,49 4:42 58 0,85 4:15
11 0,61 4:33 23 0,44 4:23 35 0,71 4:31 47 0,50 4:21 59 0,86 4:22
12 0,62 4:32 24 0,46 4:12 36 0,80 4:12 48 0,50 4:23 60 0,81 4:42
150 Izdebski, M.,
Archives of Transport, 67(3), 139-153, 2023

Table 6. Efficiency of the ant algorithm in assigning vehicles to tasks (P – probability, E – efficiency)
Test P E Test P E Test P E Test P E Test P E
1 0,44 Yes 21 0,46 Yes 41 0,49 Yes 61 0,46 Yes 81 0,49 Yes
2 0,46 Yes 22 0,44 Yes 42 0,47 Yes 62 0,46 Yes 82 0,49 Yes
3 0,47 Yes 23 0,44 Yes 43 0,47 Yes 63 0,47 Yes 83 0,48 Yes
4 0,44 Yes 24 0,44 Yes 44 0,47 Yes 64 0,47 Yes 84 0,48 Yes
5 0,44 Yes 25 0,46 Yes 45 0,48 Yes 65 0,48 Yes 85 0,47 Yes
6 0,49 No 26 0,46 Yes 46 0,45 Yes 66 0,48 Yes 86 0,46 Yes
7 0,49 Yes 27 0,44 Yes 47 0,45 Yes 67 0,49 Yes 87 0,45 Yes
8 0,49 Yes 28 0,46 No 48 0,49 No 68 0,49 Yes 88 0,53 No
9 0,46 Yes 29 0,47 Yes 49 0,47 Yes 69 0,45 Yes 89 0,46 Yes
10 0,41 Yes 30 0,47 Yes 50 0,45 Yes 70 0,45 Yes 90 0,46 Yes
11 0,44 Yes 31 0,49 Yes 51 0,47 Yes 71 0,44 Yes 91 0,45 Yes
12 0,49 No 32 0,49 Yes 52 0,47 Yes 72 0,46 Yes 92 0,44 Yes
13 0,50 Yes 33 0,46 Yes 53 0,48 Yes 73 0,52 No 93 0,44 Yes
14 0,47 Yes 34 0,44 Yes 54 0,49 Yes 74 0,46 Yes 94 0,47 Yes
15 0,47 Yes 35 0,44 No 55 0,48 Yes 75 0,47 Yes 95 0,47 Yes
16 0,44 Yes 36 0,49 Yes 56 0,45 Yes 76 0,46 Yes 96 0,48 Yes
17 0,47 Yes 37 0,46 Yes 57 0,49 Yes 77 0,47 Yes 97 0,45 Yes
18 0,45 Yes 38 0,48 Yes 58 0,48 Yes 78 0,45 Yes 98 0,45 Yes
19 0,46 Yes 39 0,48 Yes 59 0,48 Yes 79 0,48 Yes 99 0,45 Yes
20 0,44 Yes 40 0,47 Yes 60 0,47 Yes 80 0,48 Yes 100 0,48 Yes

Fig. 3. The assignment vehicles to tasks

6. Conclusions theoretical probability distributions. Using the STA-

The paper presents a new approach to risk manage- TISTICA 13 package, the distributions at each of the
ment in allocating vehicles to tasks in a transport measurement points presented in the example were
company using the ant algorithm. Dangerous events determined. A new approach in the study of the risk
on the route of vehicles have been described with of accidents in the assignment of vehicles to tasks
Izdebski, M., 151
Archives of Transport, 67(3), 139-153, 2023

was the introduction of a random variable describing References

the distribution of the probability of accidents as a [1] Agrawal, A. K., Yadav, S., Gupta A.A., & Pan-
variable interpreting the moment of the vehicle's ap- dey, S. (2022). A genetic algorithm model for
pearance at a given point on the route. optimizing vehicle routing problems with per-
The decision model considers the limitations typical ishable products under time-window and qual-
of the classic model of assigning vehicles to tasks, ity requirements. Decision Analytics Journal, 5,
e.g. window limits and additionally contains limita- 100139. [Link]
tions on the acceptable risk on the route of vehicles' jour.2022.100139
travel. The criterion function minimizes the proba- [2] Batarliene, N. (2020). Essential Safety Factors
bility of an accident occurring along the entire as- for the Transport of Dangerous Goods by Road:
signment route. A Case Study of Lithuania. Sustainability,
The proposed risk management procedure based on 12(12), 4954.
a heuristic algorithm reduces the risk to a minimum. [Link]
The ant algorithm reduces it in the event of exceed- [3] Dhouib, S. (2022). An Intelligent Assignment
ing the limit, which differs from the classic methods Problem Using Novel Heuristic: The Dhouib-
of risk management, which are dedicated only to risk Matrix-AP1 (DM-AP1): Novel Method for As-
assessment. signment Problem. International Journal of In-
The paper analyses the algorithm's sensitivity to telligent Systems and Applications in Engineer-
changes in its input parameters. Sixty combinations ing, 10(1), 135-141.
of these parameters were tested, and the combination [Link]
with the best result generated by the algorithm was [4] Ebrahim, S., Milon, I., & Quazi S.H. (2021). A
selected. It should be emphasized that the ant algo- review on neural network techniques for the
rithm is probabilistic, so the result is suboptimal. prediction of road traffic accident severity.
The proposed ant algorithm was 95% effective in as- Asian Transport Studies, 7, 100040.
sessing the risk of adverse events in assigning vehi- [Link]
cles to tasks. The algorithm was run 100 times. The [5] Fornalchyk, Y., Afonin, M., Postranskyy, T., &
designated routes were compared with the actual Boikiv, M. (2021). Risk assessment during the
hours of the accident at the bottom of the measure- transportation of dangerous goods considering
ment points. the functional state of the driver. Transport
In the scope of further work on the problem of risk Problems, 16(1), 139-152.
management in the allocation of vehicles to tasks in [Link]
transport companies, the effectiveness of other opti- [6] Fuentes, M., Cadarso, L., Vaze, V., & Barnhart,
mization algorithms, e.g. the genetic algorithm, C. (2021). The Tail Assignment Problem: A
should be examined, and the development of multi- Case Study at Vueling Airlines. Transportation
criteria risk management models with additional cri- Research Procedia, 52, 445-452.
teria functions, e.g. cost or task execution time. [Link]
[7] Giovanni, C., Giuseppe, I., Michela, L.P., Plu-
Acknowledgment chino, A., & Ignaccolo, M. (2020). Bridging the
This research was funded by the National Science gap between weak-demand areas and public
Centre, Poland (project no. 2022/04/Y/ST8/00134). transport using an ant-colony simulation-based
Project title: Energy optimal urban logistics as a ser- optimization. Transportation Research Proce-
vice (E-Laas). Project implemented as part of the dia, 45, 234-241.
call ERA-NET Cofund Urban Accessibility and [Link]
Connectivity (EN-UAC China Call) organized by [8] Haixing, W., & Qiangian, L. (2020). Risk Anal-
JPI Urban Europe and the National Natural Science ysis and Route Optimization of Dangerous
Foundation of China (NSFC) funded from the Euro- Goods Transportation Based on the Empirical
pean Union’s Horizon 2020 research and innovation Path Set. Journal of Advanced Transportation,
programme under grant agreement No 875022. 5(1), 1-13.
[Link]
152 Izdebski, M.,
Archives of Transport, 67(3), 139-153, 2023

[9] Holeczek, N. (2019). Hazardous materials truck 103, 102422.

transportation problems: A classification and [Link]
state of the art literature review. Transportation [17] Kukulski, J., Lewczuk, K., Góra, I., & Wasiak,
Research Part D: Transport and Environment, M. (2023). Methodological aspects of risk map-
69, 305-328. ping in multimode transport systems. Eksploat-
[Link] acja i Niezawodnosc – Maintenance and Relia-
[10] Hossaina, M., Abdel-Atyb, M., Quddusc M.A., bility, 25(1), 19.
Muromachid, Y., & Sadeeke, S.N. (2019). [Link]
Real-time crash prediction models: State-of- [18] Lyu, Z., & Andrew, J.Y. (2021). Consultant As-
the-art, design pathways and ubiquitous re- signment and Routing Problem With Priority
quirements. Accident Analysis and Prevention, Matching. Computers & Industrial Engineer-
124, 66-84. ing, 151(11), 106921.
[Link] [Link]
[11] Hosseini, S.D., & Verma M. (2021). Equitable [19] Mahdi, R., Amirarsalan, Mehrara, M., & Kha-
routing of rail hazardous materials shipments led, K. (2020). Analyzing injury severity of mo-
using CVaR methodology. Computers & Oper- torcycle at-fault crashes using machine learning
ations Research, 129, 105222. techniques, decision tree and logistic regression
[Link] models. International Journal of Transporta-
[12] Huang, W., Li, L., Liu, H., Zhang, R., & Xu, M. tion Science and Technology, 9(2), 89-99.
(2021). Defense resource allocation in road [Link]
dangerous goods transportation network: A [20] Mohri, S.S., Mohammadi M., Gendreau M., Pi-
Self-Contained Girvan-Newman Algorithm rayesh M., Ghasemaghaei A., & Salehi V.
and Mean Variance Model combined approach. (2022). Hazardous material transportation
Reliability Engineering & System Safety, 215, problems: A comprehensive overview of mod-
107899. els and solution approaches. European Journal
[Link] of Operational Research, 302(1),1, 1-38.
[13] Izdebski, M. , Gołda, P., & Zawisza T. (2023). [Link]
The Use of Simulation Tools to Minimize the [21] Mokhtarimousavi, S., Anderson, J.C., Az-
Risk of Dangerous Events on the Airport izinamini A., & Hadi, M. (2020). Factors af-
Apron. W: Advanced Solutions and Practical fecting injury severity in vehicle-pedestrian
Applications in Road Traffic Engineering : con- crashes: A day-of-week analysis using random
ference proceedings / Macioszek Elżbieta, parameter ordered response models and Artifi-
Granà Anna, Sierpiński Grzegorz (red.), Lec- cial Neural Networks. International Journal of
ture Notes in Networks and Systems, 604, Transportation Science and Technolog, 9(2),
Springer, 91-107, ISBN 978-3-031-22358-7. 100-115.
[14] Ji, X., & Yong, X. (2019). Application of Ge- [Link]
netic Algorithm in Logistics Path Optimiza- [22] Mujalli, R.O., Al-Masaeid, H. & Alamoush, S.
tion. Academic Journal of Computing & Infor- (2023). Modeling Traffic Crashes on Rural and
mation Science, 2, 155-161. Suburban Highways Using Ensemble Machine
[Link] Learning Methods. KSCE Journal of Civil En-
[15] Jia, Z., Yu, J., Ai, X., Xu, X., & Yang, D. gineering, 27, 814-825.
(2018). Cooperative multiple task assignment [Link]
problem with stochastic velocities and time [23] Munapo, E. (2020). Development of an accel-
windows for heterogeneous unmanned aerial erating Hungarian method for assignment prob-
vehicles using a genetic algorithm. Aerospace lems. Eastern-European Journal of Enterprise
Science and Technology, 76, 112-125. Technologies, 4(4), 6-13.
[Link] [Link]
[16] Karsu, Ö., Azizoğlu, M., & Alanl, K. (2021). 4061.2020.209172
Exact and Heuristic Solution Approaches for [24] Ongcunaruk, W., Ongkunaruk, P., & Janssens,
the Airport Gate Assignment Problem. Omega, G.K. (2021). Genetic algorithm for a delivery
Izdebski, M., 153
Archives of Transport, 67(3), 139-153, 2023

problem with mixed time windows. Computers of accident risk with transshipment option. In-
& Industrial Engineering, 159(1), 107478. ternational Journal of Production Economics,
[Link] 209, 302-315.
[25] Semenov, I., & Jacyna, M. (2022). The synthe- [Link]
sis model as a planning tool for effective supply [30] Wei, Q., Yan-Ning, S., Zi-Long, Z., Zhi-Yao,
chains resistant to adverse events. Eksploatacja L., & Yao-Ming Z. (2021). Multi-agent rein-
i Niezawodnosc – Maintenance and Reliability, forcement learning-based dynamic task assign-
24 (1), 140-152. ment for vehicles in urban transportation sys-
[Link] tem. International Journal of Production Eco-
[26] Stojanovic, N., Boskovic B., Petrovic M., Gru- nomics, 240, 108251.
jic I., & Abdullah, O.I. (2023). The impact of [Link]
accidents during the transport of dangerous [31] Yu, J., Zhitao H., Zhenyu L., & Honghai Z.
good, on people, the environment, and infra- (2023). Optimization of multi-objective airport
structure and measures for their reduction: a re- gate assignment problem: considering fairness
view. Environmental Science and Pollution Re- between airlines. Transportmetrica B:
search, 30, 32288-32300. Transport Dynamics, 11(1), 196-
[Link] [Link]://[Link]/10.1080/21680566.2022.20
[27] Szaciłło, L., Jacyna, M., Szczepański, E., & Iz- 56542
debski, M. (2021). Risk assessment for rail fre- [32] Zabielska, A, Jacyna, M., Lasota, M., & Neh-
ight transport operations, Eksploatacja i Nieza- ring. K. (2023). Evaluation of the efficiency of
wodność, Polskie Naukowo-Techniczne Towa- the delivery process in the technical object of
rzystwo Eksploatacyjne, 23(3), 476-488. transport infrastructure with the application of
[Link] a simulation model. Eksploatacja i Niezawod-
[28] Tian, Q., Li, J., Huang, G., & Yuan, W. (2022). nosc – Maintenance and Reliability, 25(1), 1.
Solving an airport ground service task assign- [Link]
ment problem with an exact algorithm. PLoS
ONE, 17(12): e0279131.
[Link]
[29] Timajchi, A., Al-e-Hashem, S.M.M., & Rekik
Y. (2019). Inventory routing problem for haz-
ardous and deteriorating items in the presence