Simulation and analysis of peer-to-peer robot swarm network
In the paper we describe the full cycle development of TDMA-based communication protocol for a swarm of robots, including simulation and hard-ware implementation. The protocol is targeted to have a robust, inter ference-immune transport in the network. First, we developed a simulator, based on SimPy...
Збережено в:
| Дата: | 2026 |
|---|---|
| Автори: | , |
| Формат: | Стаття |
| Мова: | Англійська |
| Опубліковано: |
PROBLEMS IN PROGRAMMING
2026
|
| Теми: | |
| Онлайн доступ: | https://pp.isofts.kiev.ua/index.php/ojs1/article/view/1026 |
| Теги: |
Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
| Назва журналу: | Problems in programming |
| Завантажити файл: |  |
Репозитарії
Problems in programming| _version_ | 1869381590485827584 |
|---|---|
| author | Rahozin, D.V. Smirnov, V.Ye. |
| author_facet | Rahozin, D.V. Smirnov, V.Ye. |
| author_institution_txt_mv | [
{
"author": "D.V. Rahozin",
"institution": "Institute of Software Systems NAS of Ukraine"
},
{
"author": "V.Ye. Smirnov",
"institution": "Institute of Software Systems NAS of Ukraine"
}
] |
| author_sort | Rahozin, D.V. |
| baseUrl_str | https://pp.isofts.kiev.ua/index.php/ojs1/oai |
| collection | OJS |
| datestamp_date | 2026-06-29T10:43:50Z |
| description | In the paper we describe the full cycle development of TDMA-based communication protocol for a swarm of robots, including simulation and hard-ware implementation. The protocol is targeted to have a robust, inter ference-immune transport in the network. First, we developed a simulator, based on SimPy simulation pack age, which helps us to run thousands of simulations for proving the concept of TDMA-based communications for robotic swarm. Second, using the simulator we developed a bunch of techniques for the TDMA transport to improve network robustness and simulations allowed to gather statistics and choose the better algorithm. Third, we employed so-called AI tools to implement parts of simulator and a helper technique to convert simulation code to embedded code, approaching “digital twin” paradigm. Finally, the simulated protocols are successfully ported to hardware, which supports LoRa protocol, but not limited to LoRa physical layer. The resulting embedded code works accordingly to simulation results and gathered statistics. The developed sim ulation environment and modern so-called AI tools allowed to shorten dramatically the embedded software development cycle and evaluate algorithm efficiency information from the simulation results before applying on real hardware.Prombles in programming 2026; 2: 58-66 |
| first_indexed | 2026-06-30T01:00:10Z |
| format | Article |
| fulltext |
Комп’ютерне моделювання
58
© Д.В.Рагозін, В.Є.Смірнов, 2026
ISSN 1727-4907. Проблеми програмування. 2026. №2
https://pp.isofts.kiev.ua
CC BY 4.0
УДК 681.3 https://doi.org/10.15407/pp2026.02.058
Д.В. Рагозін, В.Є. Смірнов
SIMULATION AND ANALYSIS OF PEER-TO-PEER ROBOT
SWARM NETWORK
In the paper we describe the full cycle development of TDMA-based communication protocol for a swarm of
robots, including simulation and hard-ware implementation. The protocol is targeted to have a robust, inter-
ference-immune transport in the network. First, we developed a simulator, based on SimPy simulation pack-
age, which helps us to run thousands of simulations for proving the concept of TDMA-based communications
for robotic swarm. Second, using the simulator we developed a bunch of techniques for the TDMA transport
to improve network robustness and simulations allowed to gather statistics and choose the better algorithm.
Third, we employed so-called AI tools to implement parts of simulator and a helper technique to convert
simulation code to embedded code, approaching “digital twin” paradigm. Finally, the simulated protocols are
successfully ported to hardware, which supports LoRa protocol, but not limited to LoRa physical layer. The
resulting embedded code works accordingly to simulation results and gathered statistics. The developed sim-
ulation environment and modern so-called AI tools allowed to shorten dramatically the embedded software
development cycle and evaluate algorithm efficiency information from the simulation results before applying
on real hardware.
Keywords: swarm simulation, wireless network, digital twin, TDMA-based communication
D.V. Rahozin, V.Ye. Smirnov
МОДЕЛЮВАННЯ І АНАЛІЗ РІВНОПРАВНИХ РОЇВ РОБОТІВ
У статті розглядається повний цикл розробки комунікаційного протоколу на основі TDMA для рою
роботів, включно з моделюванням та апаратною реалізацією. Протокол розроблений для забезпечення
надійного, захищеного від перешкод передавання даних у мережі. 1) Було розроблено симулятор на
основі пакета моделювання SimPy, який дозволяє проводити тисячі симуляцій для підтвердження кон-
цепції комунікацій на основі TDMA для рою роботів. 2) За допомогою симулятора розроблено низку
методів для передавання даних в рамках фреймів TDMA з метою покращення стійкості мережі. Ці
симуляції дозволили зібрати статистику, проаналізувати та вибрати кращий алгоритм. 3) Використано
інструменти так званого генеративного штучного інтелекту для створення частин симулятора та дода-
ткові техніки для перетворення коду моделювання у вбудований код з метою наближення до паради-
гми «цифрового двійника». 4) Змодельовані протоколи успішно перенесені на обладнання, яке підтри-
мує протокол LoRa, але не обмежується фізичним рівнем LoRa. Отриманий вбудований код працює
відповідно до результатів моделювання та зібраної статистики на основі моделі. Розроблене середо-
вище моделювання та сучасні інструменти штучного інтелекту дозволили значно скоротити цикл роз-
робки вбудованого програмного забезпечення та оцінити інформацію про ефективність алгоритму з
результатів моделювання перед застосуванням на реальному обладнанні.
Ключові слова: моделювання рою, бездротова мережа, цифровий двійник, зв'язок на основі технології
TDMA
1. Introduction
Today the swarm of robots or drones is
an in-demand technology for field application,
as it provides control of multiple robotic de-
vices using only one control center or even al-
low them to act autonomously following some
scenario. For various industrial purposes the
basic scenario, when an operator controls only
one drone, looks obsolete, as this case requires
a dedicated operator for each active drone. The
multiple drones use may improve the execu-
tion efficiency of many process types, but it re-
quires: 1) big enough number of operators; 2)
dedicated control channels which should be
separated one from another; 3) maybe the
strongest issue - synchronization and control-
ling operators to execute meaningful tasks over
some area without interfering one another. An-
yway, the human operator use now is the sim-
plest and the cheapest solution for many cases,
as the operator job can be done remotely for
Комп’ютерне моделювання
59
low price. But this benefit cannot be projected
for near future. The main goal of our research
is to discover the possibilities of building semi-
autonomous robot swarms, which act over de-
fined area in peer-to-peer network and may ex-
change roles under changing environmental
conditions and under limitation of robotic re-
sources until the mission is completed. One of
the main limitations is the definition of com-
munication protocol, which further should sup-
port efficient swarm control protocol.
There are many aspects that restrict the
data exchange paradigm in a robotic network,
but the first step is the building of inter-robot-
ics communication. We are not going to write
down a long list or a large classification of fac-
tors that affect the network protocol, but we
highlight the aspects most important for our
case. One aspect is the limited bandwidth, at
least for the autonomous drone scenario usu-
ally. We have no requirement for continuous
video streaming delivered for human operator,
so usually the geographical coordinates, veloc-
ity and RSSI are required data to exchange.
Another aspect is the maintaining robot net-
work integrity in case of obstacles, jamming
and noise in communication channel. The data
bandwidth aspect should be considered for es-
tablishing tradeoff between being narrow
enough against the increased number of drones
in the swarm, up to hundreds. All the aspects
restrict the tradeoffs in definition of efficient
control protocol, which allows the robotic net-
work to reach mission goals with or without
operator control. Possibly, some attention
should be paid for limiting power consumption
– to extend the life of battery-operated drones
in “suspend mode” or temporary inactive robot
mode. This enables long missions; even
weekly mission becomes possible. The proper
communication protocol gives a good basis for
building a network - with wide variety of un-
derlaying physical and transport layers. Robust
protocol is a good base for implementing par-
ticular algorithms which automatize robot mis-
sion planning and execution.
In chapter 2 we are defining the task
and goals for the robotic network concept and
discussing its most important properties. In
chapter 3 we researching the possible ways to
build a model, and describe the model of com-
munication protocol. In chapter 4 we describe
simulation results, including gathered metrics
for model efficiency and corresponding hard-
ware implementation.
2. Robotic Network Concept
2.1. Protocol concept
Our goal is to define the communica-
tion model of robotic network with respect to
modern hardware we use further for its imple-
mentation. The model is used to evaluate im-
portant metrics: 1) swarm recovery time during
mission execution in case of jamming/bad link
cases; 2) drone swarm reaction time for re-
building a swarm control software for a new
mission; 3) minimum required bandwidth for
tasks; 4) much simpler characteristics of the
swarm behavior in case of different physical
radio transceivers and power supply character-
istics. On the next step the developed model al-
lows to evaluate high-level swarm mission sce-
narios.
It should be noted, that our research
echoes the method applicable for ad-hoc net-
works [1], however the modern robots have
less limitations – more energy, less operating
time, less restrictions for transmission chan-
nels and speeds, less number of operating de-
vices. Still, the robots can move across the net-
work area, compared to practically non-mov-
ing sensor. And the main difference – if the
earlier sensors devices have quite simple on-
board sensors, modern robots have RGB and
thermal cameras, radars, lidars, ultrasonic sen-
sors and many other devices. Such device set
give the overwhelming information about en-
vironment, but still the scenarios of the robot
swarm use are quite basic, usually restricting
useful scenarios for basic operations, for ex-
ample, in agricultural sector. We are hopeful
that introducing even the basic swarm usage
scenarios into industry will help to employ
more and more use cases over near time.
For the definition of the protocol con-
cept, we are reviewing several most often used
scenarios for the swarm: 1) surveillance sce-
nario, where the swarm is constantly looking
for some anomalies over an area; 2) continuous
execution of basic tasks for robots – e. g. spray-
ing some agriculture; 3) delivering packages
over routes – in case of emergency situation –
with possible mission changes “on-the-fly”.
Комп’ютерне моделювання
60
All enlisted tasks require communication pro-
tocol for swarm orchestration, when all swarm
devices share the same quite narrow bandwidth
re-source; and a protocol for mission control.
For this paper we concentrate on communica-
tion protocol: for OSI network layers, we are
considering levels 2-3 – Dala Link and Net-
work layers, and partially level 4 – Transport
layer. The physical layer for practical evalua-
tion is fixed and the most valuable for today
are LoRa standard physical layer.
Practical considerations for scenarios
show us, that the communication layer should
concentrate at least on the following tasks: 1)
providing stable communication between the
operator and the swarm; 2) stable communica-
tion in case if operator is off-line and the net-
work is autonomous; 3) robustness in case if
some nodes fall off-line for short or moderate
time; 4) adding new nodes for swarm; 5) join-
ing two separate swarms. Looking back to pre-
viously elaborated metrics: coverage, fault tol-
erance, response time, scalability, throughput
[2], robustness, task completion rate [3], preci-
sion, success rate, adaptability to scale [4] and
so on – we are setting narrower but more com-
plex metrics and goal to simulate, as our sce-
nario set and selected physical link layer mark
several metrics as more important. So, we de-
fine a set of application-level metrics of our in-
terest, which reflect the components of swarm
mission success.
2.2. Physical layer
The most common off-the-shelf com-
munication solution is based on LoRa [5],
which looks to be well-known low-power so-
lution for environmental sensors, designed for
the long range – up to 15 km distance – appli-
cations. On the other side the communication
speed looks to be quite low, 0.3 – 50 kBit/sec,
but this does not look as a big issue. Earlier we
have discussed, that low automation degree of
drones requires high bandwidth, as the human
operator requires good quality video stream for
effective operation, for example current FPV
drones` infrastructure is built exactly such
way. If we drop video stream, we also drop
high bandwidth requirements, moving exact
object recognition operations to drone side.
This allows to fit communication requirements
into strict LoRa band-width, limiting commu-
nication to simple exchange of sensor values.
The types of sensor samples are: current GPS
position, velocity vector, power level, payload
weights or value. Basically, the required sensor
value types depend on control protocols, which
are defined on higher protocol levels, and this
requires some iterative process of defining
over-the-network control algorithms. We are
going to cover several control protocols sam-
ples in the next articles, now we concentrate on
employing off-the-shelf communication solu-
tions into our swarm.
2.3. Common considerations for
communication
Generally, the control protocol does
not depend on particular physical layer, but the
practical considerations usually point to the
cheapest “industry standard” radio transceivers
available on market. For the time of writing
this article different LoRa devices and mod-
ules, working in 433MHz non-licensed range,
are the most suitable type of devices available
on market. This does not mean that TDMA-
based protocol requires LoRa, it can be imple-
mented on any type of radio transmitting de-
vices, where the user can directly control the
transmission speed and operating modes of
transmitter. For communication protocol plan-
ning we should know basic timing delays for
switching the transmitter between generally
idle mode, receiver mode and transmitter
mode. The maximum time necessary for
changing transmitter operating mode (e.g.
from receive mode to transmission) specifies
the time gap between TDMA slots. Other point
is the accuracy of local clock, which also af-
fects time gaps between TDMA slots in proto-
col. The theoretical and practical considera-
tions for time synchronization between con-
nected devices in our network were earlier
made in [6], including practical results and
hardware implementation. The useful feature
is RSSI, which allows to evaluate the received
signal strength, so we are able to evaluate the
distance between devices and possible device
movement.
It should be noted, that robotic commu-
nication networks are developed having in
mind the target structure of the network and
use scenarios. The initial use scenarios for
swarm define the complexity of the network
Комп’ютерне моделювання
61
physical layer protocol. Basically, we can sep-
arate the network types into the following large
groups: 1) permanent configuration, where ro-
bots practically are not moved; 2) permanent
configuration where robots can move within
the limits of their defined areas, but leaving the
communications between neighbors practi-
cally unchanged – with minor distance change;
3) configuration where robots can freely move
across area. The corner cases of the last type is
the subdivision of the swarm into several
swarm with communication configuration re-
build or joining several swarms into one
swarm. All the network types also are chal-
lenged the problem of network nodes, that can
temporary be offline, so leaving the network
for short time and joining it after the leave. The
effective solution of corner cases greatly im-
proves the overall performance of the network,
and one of goals of our study is the modeling
of these corner cases scenario.
For our study we chose the scenario,
where the robots located in the geographical
center of the swarm can communicate to all the
robots in the swarm. For moderate number of
robots (80-200) we are employing TDMA-
based techniques, where we can effectively di-
vide time resource into the defined number of
time slots and give fixed amount of outcoming
traffic for each robot per time slot. The first
useful work, describing TDMA protocol for
mobile devices was described in [7], where
multi-hop mobile devices network concept for
low-speed communications was described.
Also, it was proved [6] that off-the-shelf and
low cost quartz resonators can provide time-
synchronization for robotic network. These
techniques enable to design various types
(multi-hop, one-hop) of networks, based on
TDMA communication layer techniques
which can employ LoRa physical layer.
2.4. LoRa communications
considerations
We start from considerations, that in
order to simplify TDMA network, we can em-
ploy peer-to-peer network concept over a com-
paratively large geographical area, as maxi-
mum distance for LoRa communication is up
to 15 km. One of the algorithmic improve-
ments we can employ for this case – the selec-
tion of central zone (fig. 1), where the devices
can reach all the devices in the network, but
other devices possibly cannot communicate di-
rectly to peers located too far from them. Fig.
1 shows the robots located on the semi-rectan-
gular field, the central zone devices (black
squares) can communicate to all the devices in
the network and, at least, one of them has a
kind of bridge to Internet network for reporting
swarm statistics and receiving control com-
mands. Fig. 1 shows additionally 2 devices,
marked with 1 and 2, and border limits –
dashed lines, 1-limit and 2-limit, which shows
the communication range limits for devices,
marked 1 and 2. So, comparing to TDMA-
based peer-to-peer communication networks,
the central zone controls the TDMA protocol
and slots. Other devices, competing for slots in
TDMA network, does not “hear” all the de-
vices.
Fig. 1. Two-zones robotic network structure
Consider the TDMA frame structure
for IoT devices, shown at fig. 2. The reader
may refer to [5] for comparing to simpler im-
plementation.
Fig. 2. TDMA frame structure
The total length Tframe of the TMDA
frame at fig. 2 is fixed, and this time may vary
from 0.5 sec to 15 sec, which depends on nec-
essary communication characteristics of the
network. Tframe shows the typical time of
spreading information in the network; for the
Комп’ютерне моделювання
62
network with central zone at fig. 1 this time
grows to 2*Tframe. For peer-to-peer network
this time looks very flexible in terms of con-
trol protocol.
The TDMA frame may include: start
slot X, which is employed for adding an addi-
tional robot or device to the network. Slots T0-
TK may be used for optimizing of the adding
new devices into the network – they extend slot
X. Slots 0-N are used for communicating be-
tween devices, so basically each device owns
one slot, optimized scenarios employ multiple
slots for one device to increase bandwidth. For
our case the value of N starts from 30-40 and
finishes near value 200. Some scenarios intro-
duce device priorities, so that high-prioritized
devices use several slots for communication.
Slots R0-RM are used for additional traffic op-
timizations.
As LoRa transmitters supports RSSI value,
which renders approximate distance between
receiver and transmitter, the robot approxi-
mately can have information about distance
between devices, GPS information allows to
track device movement and control the so-
called density of the devices over a field. This
also can be used for specifying communication
speed into separate slots and optimizing the
traffic inside network. For the simplest appli-
cations, only slot X and slots 0-N are used.
Slots Ti and Ri can be allocated inside slots 0-
N, employing different optimization tech-
niques, which are out of discussion scope now.
3. Swarm simulation concept
3.1. Metrics discussion
The proposed TDMA technique looks
basic, but there are several algorithm parts,
which are considered quite complex and need
to be observed and proven inside simulation.
The usual communication procedure looks
simple, but the most interesting issues are
forming the network and reentering the net-
work after communication signal jams. Com-
munication jams is a short-range or long-range
signal obstacles which prohibits communica-
tion for the part of the network. In case if the
jam is quite long in time, there is a possibility
that after such a jam the network need to build
the network from scratch. The scenarios of
joining two swarms into one or separating one
swarm are also targets for our simulation.
3.2. Simulation toolkit
The main goals of the simulation not
only for our study are at least 1) to have an en-
vironment suitable for observing and analyz-
ing the algorithms; 2) provide a cheap alterna-
tive for direct use of hardware; 3) provide a
bridge between simulated environment and
scenarios – the code in usual case – and the real
hardware. Additional point is to have ability to
switch between hardware platforms without
conceptual coder changes. It should be noted
that we are not considering computing hard-
ware of the robotic system, as communication
protocol usually requires less than 1% of over-
all system computing power.
The first common choice for such kind
of simulator is NS-3[8], or OM-NeT++ [9].
These simulators and a bunch of other simula-
tors, functionally close to it, are event-driven
simulators, which have a rich number of exten-
sions, able to simulate practically every com-
munication protocols. Still the older our work
[6] clearly shows, that for our study much
lighter tools can be involved. Our final choice
was LoRaSim[10] from Lancaster University.
Its functionality looks quite basic, but our
study showed us that this choice was right: the
simulator is lightweight and cheap in term of
resources necessary to learn it and use it. As
we use LoRa hardware as a tool, we do not
need to track much of LoRa protocol internals.
All TDMA protocols can be based on transmit-
ters, which are able to receive a packet of de-
fined length and provide a packet of defined
length. For robustness we need the ability to
change underlaying protocol without redoing
the system. So, LoRa simulator [10] is used as
an interchangeable component and a physical
layer, which can be changed to another physi-
cal layer implementation, simply connecting
channel layer in OSI model to the physical en-
vironment.
3.3. SimPy implementation
benefits
The protocol simulation is based on the
SimPy [11] – process-based discrete-event
simulation framework based on standard Py-
thon. This choice is based on its simplicity, as
Комп’ютерне моделювання
63
1) standard Python infrastructure is used; 2)
simulation is supported by shared objects and
synchronization component libraries in
SimPy; 3) the simulation code flow in SimPy
is a Python process without specific definitions
and infrastructure, which makes the SimPy use
also cheap and simple in learning how to use
and apply.
The use of common programming lan-
guage as Python showed us many benefits for
our case. First, the simulated process algorith-
mics, which implement various TDMA-net-
work support parts, is implemented as a well-
structured programming code. We even may
utilize the concept of “digital twins”, directly
translating the simulation code into the robotic
code. Despite the fact, that Python interpreter
makes the Python code slower 10-20x times
that C++ code, the modern AI tools – such as
commonly used ChatGPT or Gemini tools - al-
low seamlessly convert Python code to C++
code. The underlying protocol concept em-
ploys statically allocated components – data
buffers, arrays, variables and so one, as the
simulated process is targeted for embedded
platforms. The AI tools use allows to over-
come the barrier, which was introduced long
time ago by using different programming lan-
guages for different kind of simulations – as all
fast simulations employ C++ optimized code,
for example ROS-2 [12]. Its simulation con-
cept is an exact “digital twin” system, versus
more easily written but more slow systems in
Python, such as SimPy. Now AI tools allow to
rewrite still with some limitations the simula-
tion code from one language to another, sim-
plifying and extending the “digital twin” con-
cept use. Sure, that AI tool now cannot deal
with optimized driver code for LoRa, but the
underlying low level code base is not the sim-
ulator part. It is separated from our simulated
algorithms with abstraction layer, so we are
able to convert simply between languages even
using AI tools. For developers, who are still
worried with automated code conversions – it
is quite possible to get to mature Python-based
model, convert it to C/C++, verify differences
between original and converted code flow, and
in case of success – move further to embedded
code.
Another important point for the SimPy
use is that the Python-based development and
process-based simulator helps to use AI code
generation tools efficiently, as it can provide
the whole algorithmic base for Python effec-
tively even if the developer of the simulation
algorithm is not a proficient Python program-
mer. Although the “digital twin” concept can-
not be used directly for our model, the common
code structure for object (robot) behavior is a
common control code with time-synchroniza-
tion primitives and yield instructions, which
allows to synchronize the object behavior with
model time. During the conversion of the
model code into the hardware-side code, only
these synchronization primitives need to be
changed into hardware-related code. However,
the underlying real-time support library can
provide the corresponding compatibility layer
with SimPy synchronization primitives. All the
other control code can be built using AI tools
and we have widely practiced to use AI tools
while prototyping the simulation code.
4. Simulation details and
results
We have implemented our simulation
ecosystem on the top of LoRaSim [10] and
SimPy [11] software packages, and imple-
mented our TDMA algorithmics using this
simulation engine. Our implementation is well
aligned with concept used in ROS and ROS-2
[12] simulation, where the robotic control soft-
ware is build using “digital twin” concept. This
concept suppose that the simulation process is
identical to control program for the real robot
or drone, so the ROS simulator provides all
necessary functionality as system libraries, in-
cluding alternate versions of flight controller.
To be complete with moving the simu-
lation results to the real hardware, Raspberry
Pi boards with LoRa transceiver E32-
433T30D was used, which is connected by
UART link. As expected, real hardware elabo-
rates simulated algorithms as expected.
4.1. Simulation engine
Despite of increased computational
complexity - as proper physics simulation for
drone engine should be computed during sim-
ulation cycle – the “digital twin” approach
works well for the modern computing hard-
ware. Our solution provides real time simula-
tion for dozens of drones, including the render-
Комп’ютерне моделювання
64
ing of drone operations using 3D-engine via
Unity, and this even emulates video stream for
opera-tor reference. We successfully ap-
proached the “digital twin” concept, except we
have the main loop implemented in Python.
The structure of SimPy-based code reflects the
structure of real-time control program for
drones, so the dropped “digital twin” paradigm
component – the same programming language
– is covered with AI tools which make transla-
tion between Python and C language. We in-
troduce some internal limitations for simula-
tion code – such as static memory allocation
and simpler code structure – and current AI
tools work well enough during program map-
ping to C language. We are not going to dis-
cuss exact over-head introduced by language
translation, this may be a separate investiga-
tion mainly in the area of program code quality
and code base management.
Due to nature of the model, we have
clear layers of the code, which simplify object
synchronization in model and coding practices.
The bottom layer includes LoRaSim simulator,
which is used mainly for physical layer of data
transmission, so the practical synchronization
of robot’s work is done by LoRaSim layer. As
discussed in previous chapter, we can change
LoRaSim layer to any other wireless simula-
tion layer, as basically we need the following
functionality: 1) packet receive; 2) packet
transmission; 3) RSSI value as a part of packet
receive; 4) setting speed and transmission
power values. Any wireless protocol simulator
that provides this functionality can be used in-
stead LoRaSim. TDMA frame forming is
based on SimPy timer functionality, which can
be easily mapped on any hardware. The
TDMA frame formation principle is inspired
by ideas from old work [6], lowering the com-
plexity of multi-hop system into two-hop net-
work or star-based network. Additionally, Lo-
RaSim is slightly modified to provide tranmis-
sion jamming and simulation programmable
packet loss.
4.2. Simulation goals and metrics
On the top of TDMA frame formation
we simulate a control program, which includes
the following main parts:
1. Initial forming of TDMA net-
work, when the robotic swarm should define
priorities, example metric – the number of
visible devices or RSSI of root node which
has link to operator. Operator link is low-
speed and transfers basic swarm statistics,
goal reaching results and passes operator
commands to swarm devices. We are not
considering radio transmitter power con-
sumption as it usually less than 1% of overall
power consumption on any flying drone. One
of modeling goals was to observe different
scenarios of network forming, allowing to
improve the speed of robot joining the swarm
by extending the functionality of slot X (fig.
2). Our model allows to improve speed of
network forming 4x-8x times, depending on
slot X length and configuration.
2. Link loss correction, which has
two cases: short time link loss for several
frames, where the robot should not leave the
network and save his slot(s) active; and long-
time link loss, when the algorithm is similar to
initial network forming. This is the most inter-
esting part, as the metric of time, necessary for
forming the TDMA-based network, is the main
metric for cases, when swarm works around
concrete buildings which shield the signal
from central drones, so the drone may accom-
plish some mission in offline, join the network
again and send gathered data to robots, which
have connection to main network.
The most interesting metric we have
analyzed was the time of the network recon-
struction, when the network structure is col-
lapsed after a long signal jam. We have applied
different optimization techniques for minimi-
zation of the number of TMDA frames neces-
sary for network reconstruction, which are not
the topic of our paper, and our simulation en-
gine is able to run thousands of network recon-
struction scenarios with the results in table 1.
Table 1.
Network reconstruction times for basic and
optimized TDMA protocol.
Number of
nodes in
network
Construction
frames for the
basic period
Construction
frames for the op-
timized period
5 7 3
12 16 6
20 25 7
40 45 13
80 87 25
Комп’ютерне моделювання
65
So, the numbers in table 1 clearly
shows that our simulator allows us to con-
struct and debug algorithms quickly, as the
brute force approach to develop some algo-
rithms just on hardware simply not working.
Also we are not using any techniques of formal
verification due to its complexity, so we need
the big number of simulations with randomiz-
ing conditions in start and randomizing jam-
ming for each network configuration. Our sim-
ulation allows us to have thousands of runs for
the network with some constant number of
nodes.
4.3. Hardware platform notes
Sure, that the first move of the simula-
tion to hardware platform introduces some
problems, related to first improper assump-
tions on timings for radio transceiver. How-
ever, after initial timings correction, the simu-
lation code runs at the platform as we expected.
We use frequency band 433MHz, available for
radio amateurs, and the limit to 0.5W output
power allows to debug the protocol without
specific restrictions.
For hard debugging cases we may use
a “sniffer” hardware, which is a unit which al-
ways receives data from our wireless network,
and logs the TDMA frame structure,
frame/packets timestamps and data, so it may
be compared with simulation log.
It should be noted, that the simulation
of protocol and its different optimized version
is the only way to make an embedded system
in appropriate time. Even if hardware platform
Raspberry Pi 4 looks developer friendly, the
protocol debugging is a nutshell, as requires
elaboration of results and its analysis after each
run. Also, it is practically impossible to pro-
vide algorithmic debugging via multiple runs,
as the communication speed in LoRa protocol
makes scenarios run time extremely long.
Also, it is quite hard to provide jamming for
LoRa in real life.
We are not mention real hardware tests
at large distances. Our previous experience
with transceivers shows that in real life the
transmission distances, signal power and
speeds closely resemble the hardware manuals,
so the operational range of our transceivers
reaches the maximum, described in manuals.
Our simulation engine uses RSSI information
and introduces random errors due to communi-
cation range, so the simulation results are sim-
ilar to the real-life scenarios.
5. Conclusion
Our research clearly shows that quite
simple simulation system can greatly speed up
the development of robotic networks with
wireless communication abilities. We devel-
oped a SimPy-based simulation, which allows
to develop and debug TDMA-based protocol
for self-organized robotic swarm based on
LoRa wireless hardware, and the protocol has
advanced abilities to reconstruct network in
case of signal jams. The developed algorithms
can be comparatively easily transformed,
sometimes with helps of current AI tools, into
embedded computers, equipped with LoRa
hardware, but not limited to LoRa. We have
not noticed important differences in function-
ing of the simulator-based system and its hard-
ware counterpart, so we are close to name these
systems as digital twins. The most important is
the shortening the development cycle of com-
munication system several times, up to our ex-
perience in embedded systems development.
References
1. Agrawal R., Faujdar N., Romero C.A.T.,
Sharma O., Abdulsahib G.M., Khalaf O.I,
Mansoor R.F., Ghoneim O.A. Classification
and comparison of ad hoc networks: A review.
// Egyptian Informatics Journal, Vol. 24, Issue
1, 2023, p. 1-25, ISSN 1110-8665,
https://doi.org/10.1016/j.eij.2022.10.004.
2. Kimon P. Valavanis, George J. Vachtsevanos.
Handbook of Unmanned Aerial Vehicles.
Springer, 2014. 3022 p.
https://doi.org/10.1007/978-90-481-9707-1
3. Mohsan, S.A.H., Othman, N.Q.H., Li, Y. et al.
Unmanned aerial vehicles (UAVs): practical
aspects, applications, open challenges, security
issues, and future trends. Intel Serv Robotics
16, 109–137 (2023).
https://doi.org/10.1007/s11370-022-00452-4
4. Mueller, M., Smith, N., Ghanem, B. (2016). A
Benchmark and Simulator for UAV Tracking.
In: Leibe, B., Matas, J., Sebe, N., Welling, M.
(eds) Computer Vision – ECCV 2016. LNCS,
vol 9905. Springer, Cham.
https://doi.org/10.1007/978-3-319-46448-
0_27
Комп’ютерне моделювання
66
5. Paredes W.D., Kaushal H., Vakilinia I.;
Prodanoff Z. LoRa Technology in Flying Ad
Hoc Networks: A Survey of Challenges and
Open Issues. Sensors 2023, 23, 2403.
https://doi.org/10.3390/s23052403
6. Rahozin D. (2008) Modeling synchronised
sensor networks // Problems in Programming,
special issue, 2008, #2-3. P. 721-729.
https://nasplib.isofts.kiev.ua/han-
dle/123456789/2155
7. Kanzaki A., Uemukai T., Hara T., Nishio S..
Dynamic TDMA Slot Assignment in Ad Hoc
Networks // In Proc. 17th Int. Conf. on Ad-
vanced Information Networking and Applica-
tions (AINA’03), 2003. – P. 330–336.
8. Campanile L., Gribaudo M., Iacono M.,
Marulli F., Mastroianni, M. Computer Net-
work Simulation with ns-3: A Systematic Lit-
erature Review. Electronics, 9(2), 2020, p. 272.
https://doi.org/10.3390/electronics9020272
9. Virdis A., Kirsche M. Recent Advances in Net-
work Simulation The OMNeT++ Environment
and its Ecosystem: The OMNeT++ Environ-
ment and its Ecosystem, 2019. ISBN 978-3-
030-12841-8. https://doi.org/10.1007/978-3-
030-12842-5.
10. Voigt T., Bor M., Roedig U. Alonso, J. Miti-
gating Inter-Network Interference in LoRa
Networks. In EWSN ’17 Procs. of the 2017 Int.
Conf. on Embedded Wireless Systems and
Networks (pp. 323-328). ACM Press.
http://dl.acm.org/citation.cfm?id=3105395
11. SimPy homepage,
https://simpy.readthedocs.io/en/latest/, last ac-
cessed 2025/08/02.
12. Haridevan A., Kang J., Yuan M., Shan J.
ROS2-Gazebo Simulator for Drone Applica-
tions. In Proc. of 2024 Intl. Conf. on Un-
manned Aircraft Systems (ICUAS), p. 1232-
1238.
https://doi.org/10.1109/ICUAS60882.2024.10
556903.
Дата першого надходження до видання:
16.04.2026
Внутрішня рецензія отримана:02.05.2026
Зовнішня рецензія отримана:08.05.2026
Дата рекомендації до друку: 05.06.2026
Дата публікації: 29.06.2026
Про авторів:
Рагозін Дмитро Васильович,
кандидат технічних наук,
старший науковий співробітник
Ragozin Dmytro,
Ph.D. (technical sciences),
senior researcher
http://orcid.org/0000-0002-8445-9921.
Смірнов Валентин Євгенович,
аспірант
Smirnov Valentyn,
post-graduate student
http://orcid.org/0009-0006-4022-5951.
Місце роботи авторів:
Інститут програмних систем
НАН України
Institute of Software Systems of the
National Academy of Sciences of Ukraine
тел. +38-044-522-62-42
E-mail: ukrprog@isofts.kiev.ua
www.iss.nas.gov.ua
|
| id | pp_isofts_kiev_ua-article-1026 |
| institution | Problems in programming |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2026-06-30T01:00:10Z |
| publishDate | 2026 |
| publisher | PROBLEMS IN PROGRAMMING |
| record_format | ojs |
| resource_txt_mv | ppisoftskievua/26/7ebc7dff39e3fe3515644ce22c69ce26.pdf |
| spelling | pp_isofts_kiev_ua-article-10262026-06-29T10:43:50Z Simulation and analysis of peer-to-peer robot swarm network Моделювання і аналіз рівноправних роїв роботів Rahozin, D.V. Smirnov, V.Ye. swarm simulation; wireless network; digital twin; TDMA-based communication UDC 681.3 моделювання рою; бездротова мережа; цифровий двійник; зв'язок на основі технології TDMA УДК 681.3 In the paper we describe the full cycle development of TDMA-based communication protocol for a swarm of robots, including simulation and hard-ware implementation. The protocol is targeted to have a robust, inter ference-immune transport in the network. First, we developed a simulator, based on SimPy simulation pack age, which helps us to run thousands of simulations for proving the concept of TDMA-based communications for robotic swarm. Second, using the simulator we developed a bunch of techniques for the TDMA transport to improve network robustness and simulations allowed to gather statistics and choose the better algorithm. Third, we employed so-called AI tools to implement parts of simulator and a helper technique to convert simulation code to embedded code, approaching “digital twin” paradigm. Finally, the simulated protocols are successfully ported to hardware, which supports LoRa protocol, but not limited to LoRa physical layer. The resulting embedded code works accordingly to simulation results and gathered statistics. The developed sim ulation environment and modern so-called AI tools allowed to shorten dramatically the embedded software development cycle and evaluate algorithm efficiency information from the simulation results before applying on real hardware.Prombles in programming 2026; 2: 58-66 У статті розглядається повний цикл розробки комунікаційного протоколу на основі TDMA для рою роботів, включно з моделюванням та апаратною реалізацією. Протокол розроблений для забезпечення надійного, захищеного від перешкод передавання даних у мережі. 1) Було розроблено симулятор на основі пакета моделювання SimPy, який дозволяє проводити тисячі симуляцій для підтвердження кон цепції комунікацій на основі TDMA для рою роботів. 2) За допомогою симулятора розроблено низку методів для передавання даних в рамках фреймів TDMA з метою покращення стійкості мережі. Ці симуляції дозволили зібрати статистику, проаналізувати та вибрати кращий алгоритм. 3) Використано інструменти так званого генеративного штучного інтелекту для створення частин симулятора та дода ткові техніки для перетворення коду моделювання у вбудований код з метою наближення до паради гми «цифрового двійника». 4) Змодельовані протоколи успішно перенесені на обладнання, яке підтри мує протокол LoRa, але не обмежується фізичним рівнем LoRa. Отриманий вбудований код працює відповідно до результатів моделювання та зібраної статистики на основі моделі. Розроблене середо вище моделювання та сучасні інструменти штучного інтелекту дозволили значно скоротити цикл роз робки вбудованого програмного забезпечення та оцінити інформацію про ефективність алгоритму з результатів моделювання перед застосуванням на реальному обладнанні.Prombles in programming 2026; 2: 58-66 PROBLEMS IN PROGRAMMING ПРОБЛЕМЫ ПРОГРАММИРОВАНИЯ ПРОБЛЕМИ ПРОГРАМУВАННЯ 2026-06-29 Article Article application/pdf https://pp.isofts.kiev.ua/index.php/ojs1/article/view/1026 PROBLEMS IN PROGRAMMING; No 2 (2026); 58-66 ПРОБЛЕМЫ ПРОГРАММИРОВАНИЯ; No 2 (2026); 58-66 ПРОБЛЕМИ ПРОГРАМУВАННЯ; No 2 (2026); 58-66 1727-4907 en https://pp.isofts.kiev.ua/index.php/ojs1/article/view/1026/1094 Copyright (c) 2026 PROBLEMS IN PROGRAMMING |
| spellingShingle | swarm simulation wireless network digital twin TDMA-based communication UDC 681.3 Rahozin, D.V. Smirnov, V.Ye. Simulation and analysis of peer-to-peer robot swarm network |
| title | Simulation and analysis of peer-to-peer robot swarm network |
| title_alt | Моделювання і аналіз рівноправних роїв роботів |
| title_full | Simulation and analysis of peer-to-peer robot swarm network |
| title_fullStr | Simulation and analysis of peer-to-peer robot swarm network |
| title_full_unstemmed | Simulation and analysis of peer-to-peer robot swarm network |
| title_short | Simulation and analysis of peer-to-peer robot swarm network |
| title_sort | simulation and analysis of peer-to-peer robot swarm network |
| topic | swarm simulation wireless network digital twin TDMA-based communication UDC 681.3 |
| topic_facet | swarm simulation wireless network digital twin TDMA-based communication UDC 681.3 моделювання рою бездротова мережа цифровий двійник зв'язок на основі технології TDMA УДК 681.3 |
| url | https://pp.isofts.kiev.ua/index.php/ojs1/article/view/1026 |
| work_keys_str_mv | AT rahozindv simulationandanalysisofpeertopeerrobotswarmnetwork AT smirnovvye simulationandanalysisofpeertopeerrobotswarmnetwork AT rahozindv modelûvannâíanalízrívnopravnihroívrobotív AT smirnovvye modelûvannâíanalízrívnopravnihroívrobotív |