M.V. Khoroshailova1, A.A. Pirogov2, A.V. Turetsky3, S.Yu. Beletskaya4

1-4 FSBEI of HE “Voronezh State Technical University” (Voronezh, Russia)

1pmv2205@mail.ru; 3 pirogov.alx@gmail.com, 3 tav7@mail.ru; 4 su_bel@mail.ru 

Problem statement. In communication systems, including the operation of unmanned aerial vehicles, the radio channel is characterized by rapidly changing conditions: signal strength, signal-to-noise ratio and Doppler frequency shift can vary significantly in flight. Traditional decoders use a fixed codebook and are not able to quickly adjust the transmission rate without retraining or replacing the model, which either leads to a loss of bandwidth or a decrease in noise immunity. Storing multiple models for different speeds is unacceptable for on-board UAV systems due to memory, weight, and power consumption limitations. Therefore, the urgent task is to create a single neural network channel decoder that provides dynamic adaptation of the codebook in real-time transmission rate without retraining the model.

Goal. Development of a neural network channel decoder with dynamic adaptation of the codebook in real-time transmission rate without retraining the model for on-board UAV communication systems and systems with limited computing and energy resources.

Results. Development of a neural network channel decoder with dynamic adaptation of the codebook in real-time transmission rate without retraining the model for on-board UAV communication systems and systems with limited computing and energy resources.

Practical significance. The results obtained make it possible to use a single neural network decoder model for a wide range of transmission speeds, reducing the requirements for on-board memory, power consumption and computing resources of small-sized UAVs. The proposed approach provides adaptation to the changing quality of the radio channel in real time without restarting the decoder, which is in demand in a rapidly changing interference environment and variable-speed flights.

The work was carried out with the financial support of the Ministry of Science and Higher Education of the Russian Federation within the framework of the state assignment "youth laboratory" № FZGM-2024-0003/

Khoroshailova M.V., Pirogov A.A., Turetsky A.V., Beletskaya S.Yu. Deep learning for channel decoding with codebook adaptation by transmission rate // Radiotekhnika. 2026. V. 90. № 7. P. 62−66. DOI: https://doi.org/10.18127/j00338486-202607-11

1. Max Cohen, Guillaume Quispe, Sylvain Le Corff, Charles Ollion, Eric Moulines. Diffusion bridges vector quantized variational autoencoders. In International Conference on Machine Learning. PMLR. 2022. Р. 4141–4156.
2. Elgendy I.A., Zhang W., Tian Y.-C., Li K. Resource allocation and computation offloading with data security for mobile edge computing. Future Generation Computer Systems. 2019. V. 100. Р. 531–541.
3. Horoshajlova M.V. Razrabotka i realizaciya simmetrichnogo samoorganizuyushchegosya nejrosetevogo dekodera. Vestnik Voronezhskogo gosudarstvennogo tekhnicheskogo universiteta. 2020. T. 16. № 3. S. 60-64 (in Russian).
4. Bashkirov A.V., Horoshajlova M.V., Demihova A.S. Optimizaciya sistemy peredachi informacii ot bespilotnogo letatel'nogo apparata i povyshenie propusknoj sposobnosti kanala svyazi. Vestnik Voronezhskogo instituta MVD Rossii. 2023. № 4. S. 140–148 (in Russian).
5. Bashkirov A.V., Horoshajlova M.V. Algoritmy nizkoj slozhnosti dekodirovaniya i arhitektura dlya nedvoichnyh nizkoplotnostnyh kodov. Radiotekhnika. 2016. T. 80. № 6. S. 10-14 (in Russian).