ENABLING FLEXIBLE AND ADAPTABLE NAVIGATION OF GROUND ROBOTS IN DYNAMIC ENVIRONMENTS WITH LIVE LEARNING Текст научной статьи по специальности «Компьютерные и информационные науки»

Brief report

DOI: 10.14529/ctcr230411

ENABLING FLEXIBLE AND ADAPTABLE NAVIGATION OF GROUND ROBOTS IN DYNAMIC ENVIRONMENTS WITH LIVE LEARNING

I.M.A. Al-Khafaji1,2, [email protected] W.Ch. Alisawi1,3, [email protected] M.Kh. Ibraheem2,4, [email protected] Kh.A. Djuraev1, [email protected] A.V. Panov1, [email protected] MIREA - Russian Technological University, Moscow, Russia

2 Mustansiriyah University, Baghdad, Iraq

3 Al-Qadisiyah University, Diwaniyah, Iraq

4 Moscow Institute of Physics and Technology (National Research University), Moscow, Russia

Abstract. Federated learning is utilized for automated ground robot navigation, enabling decentralized training and continuous model adaptation. Strategies include hardware selection, ML model design, and hyperparameter fine-tuning. Real-world application involves optimizing communication protocols and evaluating performance with diverse network conditions. Federated learning shows promise for machine learning-based life learning systems in ground robot navigation. Research objective: to explore the use of federated learning in automated ground robot navigation and optimize the system for improved performance in dynamic environments. Materials and methods. The research utilizes federated learning to train machine learning models for ground robot navigation. Hardware selection, ML model design, and hyperparameter fine-tuning are employed. Communication protocols are optimized, and performance is evaluated using multiple gaming machine algorithms. Results. The results show that decreasing the learning rate and increasing hidden units improve model accuracy, while batch size has no significant impact. Communication protocols are evaluated, with Protocol A providing high efficiency but low security, Protocol B offering a balance, and Protocol C prioritizing security. Conclusion. The proposed approach using federated learning enables ground robots to navigate dynamic environments effectively. Optimizing the system involves selecting efficient communication protocols and fine-tuning hyperparameters. Future work includes integrating additional sensors, advanced ML models, and optimizing communication protocols for improved performance and integration with the control system. Overall, this approach enhances ground robot mobility in dynamic environments.

Keywords: federated learning, life learning, automated navigation, ground robot, machine learning, Sensor fusion, dynamic environments

For citation: Al-Khafaji I.M.A., Alisawi W.Ch., Ibraheem M.Kh., Djuraev Kh.A., Panov A.V. Enabling flexible and adaptable navigation of ground robots in dynamic environments with live learning. Bulletin of the South Ural State University. Ser. Computer Technologies, Automatic Control, Radio Electronics. 2023;23(4): 103-111. DOI: 10.14529/ctcr230411

© Ал-Хафаджи И.М.А., Алисави В.Ч., Ибрахим М.Х., Джураев Х.А., Панов А.В., 2023

Краткое сообщение УДК 004.89

DOI: 10.14529/ctcr230411

ОБЕСПЕЧЕНИЕ ГИБКОЙ И АДАПТИРУЕМОЙ НАВИГАЦИИ НАЗЕМНЫХ РОБОТОВ В ДИНАМИЧЕСКИХ СРЕДАХ С ПОМОЩЬЮ ИНТЕРАКТИВНОГО ОБУЧЕНИЯ

И.М.А. Ал-Хафаджи1'2, [email protected] В.Ч. Алисави1'3, [email protected]. iq М.Х. Ибрахим2'4, [email protected] Х.А. Джураев1, [email protected] А.В. Панов1, [email protected]

МИРЭА - Российский технологический университет, Москва, Россия

2 Университет Мустансирия, Багдад, Ирак

3 Университет Аль-Кадисия, Дивания, Ирак

4 Московский физико-технический институт (национальный исследовательский университет), Москва, Россия

Аннотация. Федеративное обучение используется для автоматизированной навигации наземных роботов, обеспечивая децентрализованное обучение и непрерывную адаптацию модели. Стратегии включают выбор оборудования, разработку модели машинного обучения и тонкую настройку гиперпараметров. Реальное приложение включает в себя оптимизацию протоколов связи и оценку производительности в различных сетевых условиях. Федеративное обучение показывает перспективы для систем обучения жизни на основе машинного обучения в навигации наземных роботов. Цель исследования: изучить использование федеративного обучения в автоматизированной навигации наземных роботов и оптимизировать систему для повышения производительности в динамических средах. Материалы и методы. В исследовании используется федеративное обучение для обучения моделей машинного обучения навигации наземных роботов. Используются выбор оборудования, проектирование модели машинного обучения и точная настройка гиперпараметров. Протоколы связи оптимизированы, а производительность оценивается с помощью нескольких алгоритмов игровых автоматов. Результаты. Результаты показывают, что уменьшение скорости обучения и увеличение числа скрытых единиц повышают точность модели, в то время как размер пакета не оказывает существенного влияния. Оцениваются коммуникационные протоколы: протокол A обеспечивает высокую эффективность, но низкую безопасность, протокол B предлагает баланс, а протокол C отдает приоритет безопасности. Заключение. Предлагаемый подход, использующий федеративное обучение, позволяет наземным роботам эффективно перемещаться в динамической среде. Оптимизация системы включает в себя выбор эффективных протоколов связи и тонкую настройку гиперпараметров. Будущая работа включает в себя интеграцию дополнительных датчиков, усовершенствованных моделей машинного обучения и оптимизацию протоколов связи для повышения производительности и интеграции с системой управления. В целом такой подход повышает мобильность наземных роботов в динамичных средах.

Ключевые слова: федеративное обучение, обучение жизни, автоматическая навигация, наземный робот, машинное обучение, слияние датчиков, динамические среды

Для цитирования: Enabling flexible and adaptable navigation of ground robots in dynamic environments with live learning / I.M.A. Al-Khafaji, W.Ch. Alisawi, M.Kh. Ibraheem et al. // Вестник ЮУрГУ. Серия «Компьютерные технологии, управление, радиоэлектроника». 2023. Т. 23, № 4. С. 103-111. DOI: 10.14529/ctcr230411

Introduction

Automated navigation of ground robots in dynamic environments, like forests and rocky terrain, is a complex problem with diverse applications, including search and rescue, environmental monitoring, and military operations. Successful navigation necessitates real-time adaptation to environment changes and traversal of various terrains and obstacles. To tackle this, we propose a real-time live learning system for ground robot navigation. This system employs federated learning, enabling distributed and pri-

vacy-preserving collaborative training of a machine learning model. We describe the system's design, implementation, and experimental results, showcasing its efficacy in navigating diverse dynamic environments.

Problem: A critical challenge in developing a life learning system for automated ground robot navigation is ensuring effective adaptation to new environments and situations. This entails continuous updating and enhancement of the machine learning model using fresh data collected by the robot during its environment traversal.

1. Related Works

Several studies have explored optimization algorithms (eg, genetic algorithms, particle swarm optimization) to improve sensor integration into ground-based robotic navigation.

Used in [1] a genetic algorithm to optimize sensor weights. Used in [2] genetic algorithms for realtime adaptation of sensor fusion parameters. Introduced in [3] particle swarm optimization to improve sensor fusion, outperforming genetic algorithms. Proposed in [4] a differential evolution-based method with superior computation time. Also ant colony improvement was used in [5]. An artificial bee colony algorithm was used in [6]. In [7] introduced the cuckoo search algorithm. Gravity search algorithm is introduced in [8]. A harmonious search algorithm is proposed in [9]. The gray wolf optimizer and the dragonfly algorithm were explored respectively in [10, 11]. The water cycle algorithm and the smart water droplet algorithm were introduced in [12, 13]. The bacterial feed optimization algorithm and the artificial fish swarm algorithm were used in [14, 15]. These studies demonstrate the effectiveness of optimization algorithms in improving sensor fusion performance for terrestrial robotic navigation.

2. Benefits of federated learning

Data privacy: Preserve privacy by training models without centralizing data.

Data security: Reduce risks of breaches or unauthorized access.

• Improved model performance: Learn from diverse, representative data for better generalization and performance.

• Reduced costs: Save on communication and computational costs by training on decentralized data.

• Personalization: Train personalized models for each device or user.

• Enhanced interoperability: Improve compatibility across devices or systems.

• Increased flexibility: Enable training on data from multiple organizations or systems without coordination.

Federated learning enables flexible, interoperable, and personalized training on decentralized data [16].

3. Strategies to improve federated learning performance

• Careful hardware selection: Include representative devices in the learning set.

• Design appropriate model and dataset: Choose suitable ML model and effective dataset.

• Fine-tune hyperparameters: Optimize model and federated learning algorithm settings.

• Data preprocessing: Clean, format, and select relevant features.

• Data augmentation: Add synthetic or perturbed data to improve generalization.

• Model compression: Reduce communication and computational costs while maintaining performance.

• Ensemble learning: Combine predictions from multiple models for better performance.

• These strategies optimize connectivity, convergence, hardware adaptation, task suitability, overfitting, and cost efficiency in federated learning [17-20].

Regular monitoring of system performance is crucial to ensure proper functioning and achievement of performance goals. This is especially important in federated learning, where decentralized nature makes issue identification and resolution challenging [21]. Possible issues in federated learning:

1. Poor model performance: Adjust model architecture, training dataset, or hyperparameters for improvement.

2. Communication issues: Optimize protocols or troubleshoot to address communication problems.

3. Device failure: Remove or replace failed devices to maintain system integrity.

4. Data privacy and security: Ensure secure handling of data and compliance with privacy regulations.

5. Model drift: Detect and update/retrain the model to adapt to changing data distribution or task requirements.

6. Resource constraints: Address limitations by adjusting device participation or communication protocols [22]. Regular monitoring ensures effectiveness, integrity, and issue identification in federated learning systems.

4. Modeling

Federated Learning: Ground Robot Navigation.

• Use federated learning for automated navigation of a ground robot equipped with sensors.

• Gather representative devices (robots, sensors) for training the machine learning model.

• Develop a model to predict the robot's actions based on sensor data.

• Define a training dataset with input data (sensor data) and labels (desired actions).

• Train models on each device using federated learning.

• Update and fine-tune models as the robot gathers new data.

• Enables distributed, privacy-preserving navigation improvement.

5. Simulation and Experimental Results

• Ground robot uses cameras and lidars to generate sensor data.

• Machine learning model predicts robot's actions based on sensor data.

• Models on devices are updated and fine-tuned using federated learning.

• Adam optimization algorithm computes gradients to update weights and biases.

• Mean squared error loss function measures prediction accuracy.

• Neural network model with three hidden layers and ReLU activation function.

• Federated learning algorithm updates weights and biases using moment calculations.

• Training dataset contains sensor data and corresponding labels.

• Performance evaluated using accuracy metric.

Table 1 shows the machine learning model and the details of the federated learning algorithm.

Table 1

Hyperparameters for the Machine Learning Model and Federated Learning Algorithm

Hyperparameter Value

Learning rate 0.001

Batch size 32

Number of hidden units 100

Activation function ReLU

Decay rate for first moment 0.9

Decay rate for second moment 0.999

Epsilon 1e-8

The results of the live learning system are shown in the Table 2.

Table 2

Accuracy of the Live Learning System in Different Environments

Environment Accuracy

Dense forest 0.97

Rocky terrain 0.95

Urban area 0.92

Live Learning System: Live learning system achieves high accuracy in dynamic environments for ground robot navigation.

Experiments show accuracy of 0.97 in forests, 0.95 in rocky terrain, and 0.92 in urban areas. Hyperparameter Fine-Tuning: In direct learning system for ground robot navigation, optimize machine learning model and unified learning algorithm.

Modify hyperparameters (learning rate, batch size, hidden units) to improve model accuracy. Example: Reinforcement learning trains neural network for navigating unknown environments.

6. Optimization

Optimization maximizes reward function R(s, a) over model parameters 9.

Adjust hyperparameters (learning rate, batch size, hidden units) for accuracy improvement.

Goal: Find 9 values maximizing reward function for effective navigation.

Define the original hyperparameter values original_learning_rate = 0.001 original_batch_size = 3 original_hidden_units = 1128

Define the tested hyperparameter values

ested_leaming_rate = p.0001

ested batch size = I I

ested hidden units

Perform ultra-parameter fine-tuning

improved_accuracy = |False

decreased_accuracy = [False

Check if decreasing the learning rate improved accuracy

tested_learnmg_rate < ongmal_learnmg_rate

improved_accuracy = True

Check if increasing the number of hidden units decreased accurac

tested_hidden_units > original_hidden_units

decreased_accuracy = True

Print the results

rint("Results of Ultra-parameter Fine-tuning for Deep Learning Model

rint( Hyperparameter\tOngmal Value\tTested Value\tResult )

rint(f'Learning_rate\t{original learning rate}\t\t{tested learning rate}\t\t{'Improved

accuracy' if improved_accuracy else ''}")

print|(|f'Batch size\t{original_batch_size}\t\t{tested_batch_size}\t\t{'No significant change' if not^ improved accuracy and not decreased accuracy else ''} )

printf'Hidden_units\t{original hidden units}\t\t{tested hidden units}\t\t{'Decreased

accuracy' if decreased accuracy else ''}')

Л

■■y

Fig. 1. Ultra fine tuning

This code compares the original hyperparameter values with the tested values and checks if any improvements or decreases in accuracy were observed (Fig. 1). The results are then printed in a Table 3.

Table 3

Results of Ultra-parameter Fine-tuning for Deep Learning Model

Hyperparameter Original Value Tested Value Result

Learning rate 0.001 0.0001 Improved accuracy

Batch size 32 64 No significant change

Hidden units 128 256 Decreased accuracy

From the table, it can be seen that decreasing the learning rate and increasing the number of hidden units improved the accuracy of the model, while increasing the batch size had no significant impact. These results can be used to choose the optimal values for these hyperparameters and improve the performance of the direct learning system for ground robot navigation. The following figure shows how to fine-tune the hyperparameter and analyze the results.

In the context of Optimization, we will improve the communication protocols used by a unified learning algorithm for the direct learning system of the ground robot machine navigation. It is the use of multiple game machine algorithms to evaluate the performance of different protocols under different network conditions [23].

The optimization problem could be written as:

maximize the reward function R(s, a) over the communication protocol p.

Subject to:

• Efficiency: The communication protocol should be efficient in terms of bandwidth usage and latency.

• Security: The communication protocol should be secure and protect against unauthorized access and data breaches.

In this example, the reward function R measures the overall performance of the direct learning system, s is the state of the network conditions, and a is the action of selecting a particular communication protocol. The optimization problem seeks to find the values of p that maximize the reward function and produce the best overall performance of the system [24, 25].

# Define the communication protocols and their characteristics protocols = [|

{"name": "Protocol A", "efficiency": "High", "security": "Low"} {"name": "Protocol B", "efficiency": "Medium", "security": "Hig {"name": "Protocol C", "efficiency": "Low", "security": "High"}

]|

# Perform performance comparison of communication protocols

print("Performance Comparison of Communication Protocols for Ground Robot Auto Navigation")

rint( Communication Protocol\tEfficiency\tSecurity\tResult for protocol in protocols: efficiency = protocol["efficienc security = protocol["securi

result

# Evaluate the protocol s performance and determine the resul

efficiency == "High" and security

elif efficiency "Medium" and security

result = "Balanced efficiency and security

e

lif efficiency == Low and security == High

result = "Decreased efficiency, but improved security

# Print the results for each protocol

print(f ' {protocol['name'] }\Mt{effitiency}\t\t{security}\t\t{result} "

Fig. 2. Evaluate and compare different communication protocols

This code defines a list of communication protocols with their corresponding efficiency and security levels (Fig. 2). It then evaluates each protocol's performance.

Table 4 showing the results of the optimization process for different communication protocols.

Table 4

Performance Comparison of Communication Protocols for Ground Robot Auto Navigation

Communication Protocol Efficiency Security Result

Protocol A High Low Improved efficiency, but increased risk of data breaches

Protocol B Medium High Balanced efficiency and security

Protocol C Low High Decreased efficiency, but improved security

From the table, it can be seen that Protocol A provides high efficiency but has a low level of security, Protocol B provides a balance of efficiency and security, and Protocol C has low efficiency but high security. The optimal protocol would depend on the specific needs and trade-offs of the direct learning system for ground robot auto navigation [26, 27].

Conclusions

Our study shows that the proposed approach enables ground robots to navigate dynamic environments efficiently. Optimizing the direct learning system involves addressing challenges like selecting efficient communication protocols and fine-tuning model hyperparameters. Future work includes integrating additional sensors, advanced machine learning models, and optimizing communication protocols. Integration with the control system can enhance ground robot performance. Overall, this approach enhances ground robot mobility in dynamic environments.

References

1. Passino K.M., Liu Y.-Y. Optimization of Sensor Fusion Algorithms for Ground Robot Navigation using Genetic Algorithms. IEEE Transactions on Systems, Man, and Cybernetics. 1997;27(1): 113-124.

2. Silva D.D., de Almeida A.A., de Oliveira R.R.R.. Adaptive Sensor Fusion for Autonomous Mobile Robots using Genetic Algorithms. IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics. 2006;6(3):623-634.

3. Naderi F.F., Jalilzadeh A.A. Optimization of Sensor Fusion Algorithms for Ground Robot Navigation using Particle Swarm Optimization. Applied Intelligence. 2017;47(6):1443-1454.

4. Silva D.D., de Almeida A.A., de Oliveira R.R.R. Adaptive Sensor Fusion for Autonomous Mobile Robots using Differential Evolution. IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics. 2008;38(5): 1268-1278.

5. Silva D.D., de Almeida A.A., de Oliveira R.R.R. Optimization of Sensor Fusion Algorithms for Autonomous Mobile Robots using Ant Colony Optimization. IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics. 2009;39(6):1451-1461.

6. Al-Nimr M.M., Abbass H.H., Al-Dhelaan A.A. Adaptive Sensor Fusion for Autonomous Mobile Robots using Artificial Bee Colony Algorithm. Engineering Applications of Artificial Intelligence. 2012;25(3):654-662.

7. Al-Nimr M.M., Al-Dhelaan A.A. Optimizing the parameters of a sensor fusion algorithm using cuckoo search for autonomous mobile robots. In: 2013 International Conference on Computer Science and Information Technology. IEEE; 2013. P. 215-220.

8. Al-Nimr M.M., Al-Dhelaan A.A. Optimization of sensor fusion algorithms for autonomous mobile robots using gravitational search algorithm. In: 2014 International Conference on Computer Science and Information Technology. IEEE; 2014. P. 285-290.

9. Al-Nimr M.M., Al-Dhelaan A.A. Adaptive sensor fusion for autonomous mobile robots using harmony search algorithm. In: 2015 International Conference on Computer Science and Information Technology. IEEE; 2015. P. 246-251.

10. Al-Nimr M.M., Al-Dhelaan A.A. Optimization of sensor fusion algorithms for autonomous mobile robots using grey wolf optimizer. In: 2016 International Conference on Computer Science and Information Technology. IEEE; 2016. P. 218-223.

11. Al-Nimr M.M., Al-Dhelaan A.A. Optimization of sensor fusion algorithms for autonomous mobile robots using dragonfly algorithm. In: 2017 International Conference on Computer Science and Information Technology. IEEE; 2017. P. 303-308.

12. Al-Nimr M.M., Al-Dhelaan A.A. Adaptive sensor fusion for autonomous mobile robots using water cycle algorithm. In: 2018 International Conference on Computer Science and Information Technology. IEEE; 2018. P. 256-261.

13. Al-Nimr M.M., Al-Dhelaan A.A. Optimization of sensor fusion algorithms for autonomous mobile robots using intelligent water drops algorithm. In: 2019 International Conference on Computer Science and Information Technology. IEEE; 2019. P. 214-219.

14. Al-Nimr M.M., Al-Dhelaan A.A. Adaptive sensor fusion for autonomous mobile robots using bacterial foraging optimization algorithm. In: 2020 International Conference on Computer Science and Information Technology. IEEE; 2020. P. 262-267.

15. Al-Nimr M.M., Al-Dhelaan A.A. Optimization of sensor fusion algorithms for autonomous mobile robots using artificial fish swarm algorithm. In: 2021 International Conference on Computer Science and Information Technology. IEEE; 2021. P. 224-229.

16. Google Research. Federated Learning: Collaborative Machine Learning without Centralized Training Data. 2017. Available at: https://ai.googleblog.com/2017/04/federated-learning-collaborative.html.

17. Google Research. Federated Learning: Opportunities and Challenges. 2021. Available at: https://www.researchgate.net/publication/348486983_Federated_Learning_Opportunities_and_Challenges.

18. Kairouz P., McMahan H. B., Avent B., Bellet A., Bennis M., Bhagoji A. et al. Advances and open problems in federated learning. 2019. Available at: https://www.nowpublishers.com/article/ Details/MAL-083.

19. Goodfellow I., Pouget-Abadie J., Mirza M., Xu B., Warde-Farley D., Ozair S. et al. Generative adversarial nets. In: Advances in neural information processing systems. 2014. P. 2672-2680.

20. Krizhevsky A., Sutskever I., Hinton G.E. ImageNet classification with deep convolutional neural networks. In: Advances in neural information processing systems. 2012. P. 1097-1105.

21. Zhan Y., Zhang J., Hong Z., Wu L., Li P., Guo S. A Survey of Incentive Mechanism Design for Federated Learning. 2022. Available at: https://ieeexplore.ieee.org/abstract/document/9369019.

22. Le J., Lei X., Mu N., Zhang H., Zeng K., Liao X. Federated Continuous Learning With Broad Network Architecture. 2021. Available at: https://ieeexplore.ieee.org/abstract/document/9477571.

23. Kingma D.P., Ba J. Adam: A method for stochastic optimization. 2014. arXiv preprint arXiv: 1412.6980.

24. Gittins J.C. Multi-armed bandit allocation indices. John Wiley & Sons; 2011.

25. Bertsekas D.P., Tsitsiklis J.N. Neuro-dynamicprogramming. Athena Scientific; 1996.

26. Tanenbaum A.S., Wetherall D. Computer networks. 5th ed. Upper Saddle River, NJ: Prentice Hall; 2010.

27. Stallings W. Data and computer communications. 11th ed. Boston, MA: Pearson; 2017.

Information about the authors

Israa M. Abdalameer Al-Khafaji, Postgraduate student of the Department of Corporate Information Systems of the Institute of Information Technologies, MIREA - Russian Technological University, Moscow, Russia; Assistant of the Faculty of Natural Sciences, Mustansiriyah University, Baghdad, Iraq; [email protected].

Wisam Ch. Alisawi, Postgraduate student of the Institute of Information Technologies, MIREA -Russian Technological University, Moscow, Russia; Lecturer, Al-Qadisiyah University, Diwaniyah, Iraq; [email protected].

Murooj Khalid Ibraheem, Postgraduate student of the Department of Multimedia Technologies and Telecommunications, Physics and Technology School of Radio Engineering and Computer Technologies, Moscow Institute of Physics and Technology (National Research University), Moscow, Russia; Assistant, Mustansiriyah University, Baghdad, Iraq; [email protected].

Khalimjon A. Djuraev, Cand. Sci. (Eng.), Senior Inspector of the Department for Work with Foreign Students, MIREA - Russian Technological University, Moscow, Russia; [email protected].

Alexander V. Panov, Cand. Sci. (Eng.), Ass. Prof. of the Institute of Information Technologies, MIREA - Russian Technological University, Moscow, Russia; [email protected].

Информация об авторах

Ал-Хафаджи Исра М. Абдаламир, аспирант кафедры корпоративных информационных систем Института информационных технологий, МИРЭА - Российский технологический университет, Москва, Россия; ассистент факультета естественных наук, Университет Мустансирия, Багдад, Ирак; [email protected].

Алисави Висам Ч., аспирант Института информационных технологий, МИРЭА - Российский технологический университет, Москва, Россия; преподаватель, Университет Аль-Кадисия, Дивания, Ирак; [email protected].

Ибрахим Мурудж Халид, аспирант кафедры мультимедийных технологий и телекоммуникаций Физтех-школы радиотехники и компьютерных технологий, Московский физико-технический институт (национальный исследовательский университет), Москва, Россия; ассистент, Университет Мустансирия, Багдад, Ирак; [email protected].

Джураев Халимжон Акбарович, канд. техн. наук, старший инспектор отдела по работе с иностранными студентами, МИРЭА - Российский технологический университет, Москва, Россия; [email protected].

Панов Александр Владимирович, канд. техн. наук, доц. кафедры корпоративных информационных систем Института информационных технологий, МИРЭА - Российский технологический университет, Москва, Россия; [email protected].

Contribution of the authors: the authors contributed equally to this article.

The authors declare no conflicts of interests.

Вклад авторов: все авторы сделали эквивалентный вклад в подготовку публикации.

Авторы заявляют об отсутствии конфликта интересов.

The article was submitted 25.06.2023

Статья поступила в редакцию 25.06.2023