Available online at www.sciencedirect.

com

ScienceDirect
ICT Express 11 (2025) 34–40
www.elsevier.com/locate/icte

DDS-P: Stochastic models based performance of IoT disaster detection

systems across multiple geographic areas✩
Israel Araújoa , Luis Guilherme Silvaa , Carlos Britoa , Dugki Minb ,∗, Jae-Woo Leec , Tuan
Anh Nguyenb,d ,∗, Erico Leãoa , Francisco A. Silvaa
a Laboratory of Applied Research to Distributed Systems (PASID), Federal University of Piaui (UFPI), Picos, Piaui 64607-670, Brazil
b Department of Artificial Intelligence, Graduate School, Konkuk University, Seoul 05029, South Korea
c Department of Mechanical and Aerospace Engineering, Konkuk University, Seoul 05029, South Korea
d Konkuk Aerospace Design-Airworthiness Research Institute (KADA), Konkuk University, Seoul 05029, South Korea

Received 26 April 2024; received in revised form 13 June 2024; accepted 8 September 2024
Available online 12 September 2024

Abstract
Effective management of catastrophic events in high-risk zones necessitates a holistic technological approach to protect ecosystems,
biodiversity, and native populations. Limitations in sensor range and connectivity hamper real-time data gathering in secluded areas, while
financial and technical hurdles hinder the creation of cost-effective, automated systems. This study presents stochastic models, the LoRaW
protocol, and cloud technology to enhance sensor deployment simulations. Wireless Sensor Networks and LoRa technology are crucial
for extensive monitoring and communication infrastructures. Stochastic Petri Net models optimize system components by assessing crucial
performance indicators, such as average response time and system utilization, thus improving disaster response and supporting research
hypotheses.
© 2024 The Author(s). Published by Elsevier B.V. on behalf of The Korean Institute of Communications and Information Sciences. This is an open
access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Keywords: Disaster monitoring; Lora; Performance; Stochastic petri nets; Sensitivity analysis

1. Introduction frameworks. LoRaWAN is a low-power, wide-area network

The urgent need to detect and manage disasters in hazard- protocol that connects battery-operated devices to the inter-
prone areas necessitates integrating advanced technologies like net. It uses Chirp Spread Spectrum (CSS) technology for
Long Range Wide Area Network (LoRaWAN) within IoT long-range communication with low power consumption. Lo-
RaWAN networks, consisting of end devices, gateways, and
✩ This research was partially supported by Basic Science Research servers, are arranged in a star topology. End devices commu-
Program through the National Research Foundation of Korea (NRF) funded nicate with gateways via single-hop wireless links, and the
by the Ministry of Education, South Korea (No. 2020R1A6A1A03046811).; gateways relay the data to the network server, enabling sig-
This research was supported by the Basic Science Research Program through
nificant power savings and extended battery life. Operating in
the National Research Foundation of Korea (NRF) funded by the Ministry
of Education (2021R1A2C2094943), South Korea. This work was carried unlicensed ISM bands, LoRaWAN is cost-effective and glob-
out with financial support from the National Council for Scientific and ally deployable, with robust security features like AES-128 en-
Technological Development (CNPq) under process PDPG-POSDOC-AUXPE cryption [1]. This protocol is ideal for monitoring vast forests,
No. 88881.830176/2023-01, and the Brazilian National Council for Scientific addressing the increasing frequency of forest fires, projected to
and Technological Development – CNPq, Universal Project 420365/2023-0, rise by 50% by 2100 [2]. Stochastic Petri Nets (SPNs) play a
Brazil;.
∗ Corresponding author at: Department of Artificial Intelligence, Graduate crucial role in modeling and analyzing the dynamic behaviors
School, Konkuk University, Seoul 05029, South Korea. of these networks to ensure operational correctness [3]. By
E-mail addresses: israel.araujo@ufpi.edu.br (I. Araújo), employing SPNs in conjunction with LoRaWAN, Wireless
luis.e@ufpi.edu.br (L.G. Silva), carlosvictor@ufpi.edu.br (C. Brito), Sensor Networks (WSNs) can enable early and precise disaster
dkmin@konkuk.ac.kr (D. Min), jwlee@konkuk.ac.kr (J.-W. Lee), detection through long-distance, low-power data transmission
anhnt2407@konkuk.ac.kr (T.A. Nguyen), ericoleao@ufpi.edu.br (E. Leão),
and simulation of probabilistic system behaviors.
faps@ufpi.edu.br (F.A. Silva).
Peer review under responsibility of The Korean Institute of Communica- This research promotes the use of stochastic modeling
tions and Information Sciences (KICS). to simulate sensor operations within a LoRaWAN-enhanced
https://doi.org/10.1016/j.icte.2024.09.005
2405-9595/© 2024 The Author(s). Published by Elsevier B.V. on behalf of The Korean Institute of Communications and Information Sciences. This is an
open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
I. Araújo, L.G. Silva, C. Brito et al. ICT Express 11 (2025) 34–40

environment, supported by fog computing for improved system real scenarios [1]. Additionally, [2] focuses on enhancing data
integration. The objective is to develop cost-effective and sus- transmission efficiency and improving responder tracking. Per-
tainable systems for disaster prevention, aimed at enhancing formance optimization: The use of Petri net-based modeling
response capabilities to protect environments, communities, for performance metrics’ optimization has been highlighted
and natural resources. The study assesses the effectiveness in other works. A SPN model was developed by Bhadra
of integrated sensors and LoRaWAN for disaster detection, et al. [3] to optimize CI/CD pipelines, using sensitivity anal-
evaluates system performance including sensor processing and ysis to enhance timely delivery. Ni et al. [10] introduced a
data transmission, and proposes the application of SPN for Priced Timed Petri Net (PTPN) model for fog computing
system simulation and optimization. It hypothesizes that the resource allocation, predicting task completion time and cost,
use of LoRaWAN and sensors for remote disaster detection is and employing a dynamic strategy to reduce response time.
feasible and efficient, focusing on server capacity planning. Analogy of this work: Our research integrates an SPN model
The key contributions of this study include: (i.) Imple- for disaster monitoring, contrasting with traditional approaches
mentation of Stochastic Models for IoT Sensor Deployment by using predictive modeling to understand system behaviors
in LoRaWAN Protocols, which simulate the deployment and across different scenarios. We evaluate the model configura-
operation of IoT sensors; (ii.) Application of SPNs for De- tions based on metrics such as MRT and Detection Probability
tailed Performance Optimization of Disaster Detection Sys- (DP), enrich our model with an absorbing state to examine
tems, enhancing monitoring accuracy and response effective- metrics like Mean Time To Absorb (MTTA), and conduct a
ness; (iii.) Comprehensive Performance Metric Analysis for sensitivity analysis using Design of Experiments (DoE).
IoT-Based Disaster Detection Systems, identifying operational
3. System architecture
efficiencies and effectiveness; (iv.) Strategic Integration of
Cloud Computing with LoRa Technology for Enhanced Data Overall Description: Fig. 1(a) depicts the network
Processing, enabling robust data management and real-time architecture designed for disaster monitoring in forested areas,
processing; (v.) Economic and Scalable System Design for connecting sensors to The Things Network (TTN), an Applica-
Wide-Area IoT-Based Disaster Detection, addressing financial tion Server, and a Monitoring Server to ensure robust surveil-
and geographic challenges in extensive disaster-prone regions. lance [11]. Covering regions with varying disaster risk levels
The findings reveal that SPN models enhance operational — high, medium, and low — the system uses LoRaWAN
efficiency and extend sensor communication range while main- technology via central access points that facilitate data com-
taining low power consumption, crucial for large, remote munication over the Internet. Class A LoRaWAN protocol is
areas. The simulation results confirm the system’s capability employed for optimal energy use, enhancing sustainability by
for swift disaster response. The implications include scal- reducing battery replacements. Sensor data are managed on
ability and cost-effectiveness of the model, potential influ- a network storage server, with TTN’s infrastructure ensuring
ence on policy-making, and technological advancement in IoT efficient data handling and integrity. The Application Server
applications for disaster management. processes and converts this data for analysis and visualization,
The paper is structured as follows: Section 2 reviews related providing customized monitoring solutions.
works. Section 3 details the system architecture. Section 4 Challenges: Key challenges include maintaining cover-
presents models with/without absorbing states and numerical age over extensive areas, optimizing sensor placement against
analyses. Section 5 concludes the paper and suggests future natural elements, and ensuring sensor durability and accu-
research directions. racy amid environmental challenges. These issues are tackled
through strategic sensor placement, protective measures, tech-
nology redundancy, and routine maintenance. LoRa Nodes:
2. Related work
LoRa nodes are essential for collecting environmental data,
The previous studies using methodologies akin to ours, adhering to the 30-30-30 rule for disaster evaluation, which
aimed at disaster prevention and detection in high-risk areas facilitates timely disaster alerts. Their effectiveness is influ-
are categorized into two groups based on their evaluation tech- enced by regional variables like rainfall, demonstrating the
niques, emphasizing the influence of performance measure- adaptability of the system. Sensors: The architecture integrates
ment choices on assessment accuracy and cost-effectiveness. various sensors such as DHT22 and MQ135 to monitor tem-
Disaster Prevention: Research such as [4] employs SPNs perature, humidity, and CO2 levels, alongside a modified DC
for disaster prevention, involving static and dynamic analysis motor for measuring wind speed, calibrated against standard
based on expert opinions. Later studies, like [5], assess risks anemometers. This diverse sensor setup enables comprehen-
in subway systems using cloud-based Petri net models com- sive environmental monitoring, vital for accurate disaster risk
bined with fault tree analysis. Another significant contribution assessment.
is [6], which optimizes disaster response strategies by sug-
gesting targeted resource allocation and evaluating emergency 4. Stochastic models
response effectiveness [7]. Disaster Detection: Recent ad- 4.1. Model without absorbing state
vancements in disaster detection include deploying unmanned
aerial vehicles (UAVs) [8], integrating IoT solutions with Model Description: Fig. 1(c) outlines the SPN model,
LoRaWAN [9], and developing preemptive systems tested in consisting of five key layers: Data Arrival, Gateway, The
35
I. Araújo, L.G. Silva, C. Brito et al. ICT Express 11 (2025) 34–40

Fig. 1. System architecture and stochastic models.

Table 1 In the SPN model depicted in Fig. 1(c), specific values are as-
Values for transitions and markings of the SPN model. signed to transitions and markings to simulate realistic system
Tad1, Tad12, Tad3 50.0 (ms) operations. Table 1 details these values, which are crucial for
Tqg1 5.0 (ms) understanding the model’s performance under various scenar-
Time Transitions Tqttn1 25.0 (ms)
Tqsa1 20.0 (ms)
ios. These values are derived from literature’s empirical data
Tqsm1 1000.0 (ms) and preliminary simulations, ensuring the model’s accuracy
Pad1, Pad2, Pad3 1.0
and relevance.
Parq1, Parq2, Parq3 0 Model Metrics: This subsection discusses the SPN
Pqg1 0 model’s metrics to evaluate system performance, focusing on
Pcqg1 20 MRT, fog processing utilization, data communication integrity,
Pcqttn1, Pcqsa1 50.0 (ms) and overall throughput. MRT, detailed in Eq. (1a), measures
Places
Pcqsm1 200.0 (ms)
Pcpttn1 50
the average time to process requests, indicative of the system’s
Pcpsa1 42 response efficiency. Utilization, defined in Eq. (1b), calculates
Pcpsm1 2 the resource usage efficiency in the fog layer by assessing
Pqttn1, Pqsa1, Pqsm1 0 the processing queue’s average token load relative to total
capacity. Discard Probability (DP), explained in Eq. (1c),
assesses the likelihood of data loss due to overloads, crucial for
Things Network (TTN) Server, Application Server, and Moni- maintaining data integrity during critical operations. Through-
toring Server. The Data Arrival layer receives system requests put (TP), shown in Eq. (1d), evaluates the system’s capacity to
from sensor anomalies and forwards them to the Gateway, handle and transfer data effectively. These metrics are instru-
which routes these requests to their destinations. The TTN mental for understanding and enhancing operational efficiency
Server processes and stores the sensor data, subsequently in disaster management scenarios, providing a foundation for
sending it to the Application Server for integration and vi- future advancements in disaster response technologies. The
sualization. The Monitoring Server provides a user interface SPN model was validated through stationary simulations with
for ongoing system surveillance. The SPN model incorporates a 2% error margin using Mercury Tool version 5.0.1.1
operational elements like Pad1, Pad2, and Pad3 for initiating
sensor requests, and queues such as Parq1, Parq2, Parq3, and 1
(
MRT = × Esp(Pqd1) + Esp(Pd1)
Pqg1 for managing data flow. Server capacities are controlled 1 − DP
through queues Pcqttn1, Pcqsa1, and Pcqsm1, while process- + Esp(Pqd2) + Esp(Pd2)
ing abilities are defined by Pcpttn1, Pcpsa1, and Pcpsm1. + Esp(Pqtc) + Esp(Pqgat) (1a)
Timed transitions like Tad1, Tad2, and Tad3 manage the + Esp(Pqe1) + Esp(Pe1) + Esp(Pqe2)
intervals between sensor requests. Transitions Tqg1, Tqttn1, )
Tqsa1, and Tqsm1 facilitate data progression through the sys- + Esp(Pe2) + Esp(Pqb) ×AD
tem. Marking places indicate the maximum capacities (C G, Esp(Pn1)
C TTN, C SA, C SM) and the available processing resources UN = × 100 (1b)
Cor eE
(N TTN, N SA, N SM) at each server, essential for accommo-
dating variable loads and ensuring efficient data management. 1 https://www.modcs.org/
36
I. Araújo, L.G. Silva, C. Brito et al. ICT Express 11 (2025) 34–40

Fig. 2. Analysis results varying processing cores and service time across two scenarios.

effectively meeting high demands characterized by high data

DP = P(Pcqtc = 0) × 100 (1c) arrival rates, low discard probabilities, and optimal response
times. This choice ensures the system’s efficiency and cost-
Esp(Pqc1) effectiveness by avoiding unnecessary hardware and energy
TP = (1d)
Ser viceT ime costs.
Case-studies: This section evaluates two scenarios Scenario 2 — Variation in Fog Service Time: The impact
within a disaster management model, focusing on the number of varying service times (100 ms to 2000 ms) on fog devices
of processing cores and service times in the fog layer. The was analyzed. Figs. 2(e)–2(h) revealed that longer service
objective is to determine the optimal configurations that im- times elevated MRTs, notably at 2000 ms, where increased
prove system efficiency. The primary goal is to optimize the arrival rates correlated with higher response times. Utilization
number of processing cores at fog nodes to enhance disaster trends (Fig. 2(f)) showed a rise in utilization in the fog layer
monitoring and response. By analyzing these setups, we aim with arrival rates, reaching 90% in scenarios with extended
to identify configurations that maximize system performance, service times. Shorter service times facilitated efficient pro-
guiding architectural improvements. System performance was cessing, alleviating stress on the fog layer. Elevated service
assessed using metrics such as MRT, Drop Probability (DP), times heightened discard probability, notably at lower arrival
Throughput (TP), and Utilization of the fog computing layer. rates during prolonged service durations, impacting system
These metrics provide insights into the system’s response time, reliability. Throughput analysis (Fig. 2(h)) indicated optimal
processing capacity, and resource usage. throughput at 100 ms service time, decreasing with extended
Scenario 1 — Variation in the Number of Fog Cores: In the service times and higher arrival rates. These findings are
first scenario, the number of fog layer cores (2, 4, 8, 16, 32) vital for optimizing fog computing configurations in disaster
was increased, and the impact on system metrics was analyzed. management systems.
Results, as shown in Figs. 2(a)–2(d), indicate that performance
is generally enhanced by more cores, especially for larger 4.2. Model with absorbing state
data volumes and real-time responses. It was revealed that 8
cores suffice for high demands, with minimal gains beyond Model Structure: Fig. 1(b) displays the model modifi-
this point. In the second scenario, 8 cores were kept constant cations necessary for integrating an efficient absorbing state.
while service time (100 ms to 2000 ms) was varied to assess Changes were primarily made to the Admission and Outbound
system efficiency. As data volume increased, the fog layer’s Processing layers to enhance the system’s capability to effi-
capacity was reached at high arrival rates, demonstrating its ciently absorb elements. These adjustments help evaluate the
load management ability. Request discards were reduced by time it takes for elements to be absorbed and the probability
higher core counts, improving capacity, but throughput was not of absorbing a specific number of elements within a given
always enhanced, indicating an optimal performance thresh- timeframe, key metrics for system evaluation. In the Admis-
old. The selection of 8 cores in the second scenario was sion layer, adjustments to places Pad1, Pad2, and Pad3 control
informed by a comprehensive analysis of performance metrics the entry of requests, with a variable KEL introduced to cap
including MRT, Utilization, DP, and Throughput. Our findings entry from these locations. For instance, if KEL is set to 10,
indicated that while increasing the number of cores to 32 no more than 10 elements can enter from each place, man-
slightly enhanced performance, the improvement beyond 8 aging the flow into the system and preventing overloads. The
cores was marginal. Specifically, 8 cores provided an opti- Outbound Processing layer includes place Pabs, an absorbing
mal balance between performance and resource utilization, place indicating whether transactions or requests have been
37
I. Araújo, L.G. Silva, C. Brito et al. ICT Express 11 (2025) 34–40

Fig. 3. Analysis results and absorption probability.

Table 2 Table 3
Design of experiments. Combination of factors considering the MRT(ms) metric.
Factor name Low setting High setting C_G C_TTN C_SA C_SM N_SM MRT(ms)
C_G 20.0 40.0 20.00 50.00 50.00 600.00 2.00 11 789.14
C_TTN 50.0 100.0 20.00 50.00 50.00 600.00 4.00 8247.83
C_SA 50.0 100.0 20.00 50.00 50.00 1200.00 2.00 12 503.13
C_SM 600.0 1200.0 20.00 50.00 50.00 1200.00 4.00 6409.96
N_SM 2.0 4.0 20.00 50.00 100.00 600.00 2.00 12 054.15
20.00 50.00 100.00 600.00 4.00 9286.06
20.00 50.00 100.00 1200.00 2.00 11 401.64
20.00 50.00 100.00 1200.00 4.00 6475.20
fully processed. This modification allows precise measurement 20.00 100.00 50.00 600.00 2.00 12 560.43
of the absorbed requests, offering insights into the system’s 20.00 100.00 50.00 600.00 4.00 7951.83
processing efficiency and capacity. The correlation of incom- 20.00 100.00 50.00 1200.00 2.00 10 872.14
ing elements (KEL) with absorption at Pabs provides valuable 20.00 100.00 50.00 1200.00 4.00 6984.48
20.00 100.00 100.00 600.00 2.00 11 816.65
data on the system’s capacity to manage and process incoming 20.00 100.00 100.00 600.00 4.00 6451.89
demands effectively. 20.00 100.00 100.00 1200.00 2.00 11 308.01
Analyses: The analysis in Fig. 3(c) evaluates system 20.00 100.00 100.00 1200.00 4.00 7781.86
performance under varying request volumes (K = 10, 20, 30) 40.00 50.00 50.00 600.00 2.00 10 984.09
40.00 50.00 50.00 600.00 4.00 7369.21
using the Cumulative Distribution Function (CDF) for comple-
40.00 50.00 50.00 1200.00 2.00 11 467.05
tion probability within specific timeframes and the Mean Total 40.00 50.00 50.00 1200.00 4.00 8150.59
Absorption Time (MTTA) for efficiency. Lower K values yield 40.00 50.00 100.00 600.00 2.00 11 622.22
higher efficiency, with completion times up to 10 000 units, 40.00 50.00 100.00 600.00 4.00 5383.04
while higher K values extend completion times beyond 20 000 40.00 50.00 100.00 1200.00 2.00 12 468.02
40.00 50.00 100.00 1200.00 4.00 6362.54
units, reducing completion probability. Scenarios (T1, T2, T3) 40.00 100.00 50.00 600.00 2.00 11 385.80
examine absorption rates over intervals of 0–5300, 5300–6200, 40.00 100.00 50.00 600.00 4.00 7681.47
and 6200–8500 h, considering arrival rates (10, 50, 100 units). 40.00 100.00 50.00 1200.00 2.00 11 477.62
T1 shows higher absorption at lower arrival rates, decreasing 40.00 100.00 50.00 1200.00 4.00 7144.52
40.00 100.00 100.00 600.00 2.00 12 135.27
with increased rates. T2 and T3 indicate that longer periods
40.00 100.00 100.00 600.00 4.00 7690.10
might enhance absorption despite higher rates. Fig. 3(a) shows 40.00 100.00 100.00 1200.00 2.00 10 991.31
that higher arrival rates increase MTTA due to longer intervals 40.00 100.00 100.00 1200.00 4.00 7487.34
between inputs. Conversely, Fig. 3(b) illustrates that more
processing cores in the fog layer significantly reduce MTTA,
highlighting core count’s importance in optimizing fog layer factors with substantial MRT impact. Table 2 lists these fac-
efficiency. tors and their settings, the values result from a factorial de-
sign, varying factors at defined levels to analyze their im-
4.3. Design of experiments pact on MRT, ensuring robust, practical optimization insights.
While, Table 3 presents their experimental combinations, aid-
This study employs SPNs to assess MRT and its impact ing in understanding how these factors alter system responsive-
on system performance, focusing on optimization opportu- ness. Analysis of Factors Affecting MRT: Effects graphs
nities. Using the Design of Experiments (DoE) methodol- (Fig. 4(a)) rank factors by their MRT influence. Key findings
ogy, we systematically analyze variable effects. By merg- indicate the Monitoring Server’s capacity as a primary deter-
ing SPN modeling with DoE, we perform a sensitivity anal- minant of performance, suggesting that enhancements here can
ysis to identify critical factors influencing system outputs. significantly improve efficiency. Other notable factors include
Our experiments manipulate variables and measure the MRT the number of Monitoring Server cores and the interaction
variations, utilizing factor interaction and effects graphs to between the gateway and Monitoring Server, which are crucial
highlight significant variables. Sensitivity analysis prioritizes for optimizing response time. The TTN Server capacity has a
38
I. Araújo, L.G. Silva, C. Brito et al. ICT Express 11 (2025) 34–40

Fig. 4. Impact and interaction of model factors on MRT; (a) Impact of model factors, (b) Gateway vs. TTN Server, (c) App Server vs. Monitoring Cores,
(d) TTN vs. App Server.

lesser impact, directing optimization efforts towards more crit- CRediT authorship contribution statement
ical factors to enhance performance. Sensitivity Analysis
Israel Araújo: Writing – original draft, Visualization,
of Impacting Factors in System Performance: Further
Validation, Software, Formal analysis, Data curation. Luis
analysis (Fig. 4(b)) demonstrates that increasing both Applica-
Guilherme Silva: Writing – original draft, Visualization,
tion Server capacity and Monitoring Server cores significantly Validation, Software, Formal analysis, Data curation. Carlos
reduces MRT. Fig. 4(c) explores interactions between TTN Brito: Writing – review & editing, Visualization, Validation,
Server and Application Server capacities, uncovering complex Software, Resources, Formal analysis. Dugki Min: Writing –
patterns that necessitate careful adjustments for optimal per- review & editing, Validation, Supervision, Investigation, Fund-
formance. Fig. 4(d) shows the effect of TTN and App Server ing acquisition. Jae-Woo Lee: Writing – review & editing,
capacities on MRT, indicating constant MRT for C S A = 50.0 Validation, Supervision, Methodology, Investigation, Funding
and increasing MRT for C S A = 100.0, suggesting inefficien- acquisition. Tuan Anh Nguyen: Writing – review & editing,
cies at higher capacities and emphasizing balanced capacity Validation, Supervision, Resources, Project administration,
planning in IoT disaster detection systems. Methodology, Investigation, Conceptualization. Erico Leão:
Writing – review & editing, Validation, Project administration,
Investigation, Formal analysis. Francisco A. Silva: Writing –
4.4. Discussions
original draft, Validation, Supervision, Project administration,
Methodology, Investigation, Conceptualization.
(i.) Machine learning and real-world testing measures: This
study uses SPNs to optimize IoT-based disaster detection by Declaration of competing interest
analyzing MRT and DP. SPNs were chosen for their precise
modeling of asynchronous activities over machine learning. The authors declare that there is no conflict of interest in
this paper.
Future work will integrate machine learning for better pre-
dictive capabilities and conduct real-world testing to validate
the model’s effectiveness in dynamic environments. (ii.) Cost- References
effectiveness analysis measures: While the study demonstrates [1] M. Antunes, L.M. Ferreira, C. Viegas, A.P. Coimbra, A.T. de Almeida,
cost-effectiveness through the use of low-power LoRaWAN, Low-cost system for early detection and deployment of countermea-
sures against wild fires, in: 2019 IEEE 5th World Forum on Internet
optimized server capacity, and fog computing, a detailed cost of Things (WF-IoT), 2019, pp. 418–423, http://dx.doi.org/10.1109/WF-
evaluation will be included in future work. This will encom- IoT.2019.8767331.
pass comprehensive cost–benefit analyses, economic impact [2] S. Debdas, A. Chakraborty, H.K. Verma, A. Kushwaha, D.V.P. Varma,
studies, and financial simulations to ensure robust economic A. Karmakar, Long range (LoRa) fire fighter in dense forest, in: 2022
International Conference on Computer, Power and Communications,
assessments. ICCPC, 2022, pp. 286–293, http://dx.doi.org/10.1109/ICCPC55978.
2022.10072080.
[3] S. Bhadra, A stochastic Petri net model of continuous integration
5. Conclusion and continuous delivery, in: 2022 IEEE International Symposium
on Software Reliability Engineering Workshops, ISSREW, 2022, pp.
The study validates SPN models for optimizing disaster 114–117, http://dx.doi.org/10.1109/ISSREW55968.2022.00050.
prevention systems, emphasizing key metrics like MRT and [4] Y. Tao, D. Xia, Y. Zhu, Research on emergency decision-making model
for major disasters based on stochastic Petri nets, in: 2019 9th Inter-
system utilization. Future research will integrate real-time
national Conference on Fire Science and Fire Protection Engineering,
data, develop adaptive algorithms, and explore machine learn- ICFSFPE, 2019, pp. 1–6, http://dx.doi.org/10.1109/ICFSFPE48751.
ing for better disaster prediction and mitigation. Additional 2019.9055778.
metrics, IoT-based real-world testing, simulation validation, [5] C. Zhang, G. Tian, A.M. Fathollahi-Fard, W. Wang, P. Wu, Z. Li,
Interval-valued intuitionistic uncertain linguistic cloud Petri net and its
and comparative analysis in disaster-prone areas will enhance
application to risk assessment for subway fire accident, IEEE Trans.
robustness and effectiveness, highlighting potential and limi- Autom. Sci. Eng. 19 (1) (2022) 163–177, http://dx.doi.org/10.1109/
tations. TASE.2020.3014907.
39
I. Araújo, L.G. Silva, C. Brito et al. ICT Express 11 (2025) 34–40

[6] J. Zhou, G. Reniers, Petri-net based simulation analysis for emer- [9] P. Luna, S. Gutiérrez, R. Espinosa, Design and implementation of
gency response to multiple simultaneous large-scale fires, J. Loss a node geolocation system for fire monitoring through LoRaWAN,
Prev. Process Ind. 40 (2016) 554–562, http://dx.doi.org/10.1016/ in: 2020 IEEE International Autumn Meeting on Power, Electronics
j.jlp.2016.01.026, URL: https://www.sciencedirect.com/science/article/ and Computing, ROPEC, 4, 2020, pp. 1–6, http://dx.doi.org/10.1109/
pii/S0950423016300249. ROPEC50909.2020.9258741.
[7] H. Ning, J. Longzhe, W. Zongzhi, X. Bo, Performance analysis [10] L. Ni, J. Zhang, C. Jiang, C. Yan, K. Yu, Resource allocation strategy
for emergency response capability based on stochastic Petri net, in fog computing based on priced timed Petri nets, IEEE Internet
in: 2010 IEEE International Conference on Emergency Management Things J. 4 (5) (2017) 1216–1228, http://dx.doi.org/10.1109/JIOT.2017.
and Management Sciences, 2010, pp. 61–64, http://dx.doi.org/10.1109/ 2709814.
ICEMMS.2010.5563503. [11] I. Afanasyev, M. Mazzara, S. Chakraborty, N. Zhuchkov, A. Mak-
[8] G. Hristov, J. Raychev, D. Kinaneva, P. Zahariev, Emerging methods satbek, A. Yesildirek, M. Kassab, S. Distefano, Towards the internet
for early detection of forest fires using unmanned aerial vehicles of robotic things: Analysis, architecture, components and challenges,
and lorawan sensor networks, in: 2018 28th EAEEIE Annual Con- in: 2019 12th International Conference on Developments in ESystems
ference, EAEEIE, 2018, pp. 1–9, http://dx.doi.org/10.1109/EAEEIE. Engineering (DeSE), IEEE, 2019, pp. 3–8.
2018.8534245.

40