CNN-Based LDPC Decoder for Hubless Full-Mesh VSATPlus ^® System

The transition of air traffic control (ATC) communication networks from traditional circuit-switched systems to modern Internet protocol (IP)-based architectures marks a significant evolution in aviation safety and efficiency. A critical component of this migration is the adoption of very small aperture terminal (VSAT) satellite communications, which enables connectivity in the areas where fiber build-out is lacking or can provide a backup to fiber connectivity. VSAT links have to overcome signal interference issues and utilize modern error correction code (ECC) techniques to minimize packet loss. The implementation of international civil aviation organization (ICAO)-recommended systems for improving traffic flow and safety modernization are built upon an IP network. However, IP protocol is built upon the central concept of low bit errors transmissions and a single bit error results in the loss of an entire IP packet. This paper investigates the integration of low-density parity-check (LDPC) decoding methods, enhanced by convolutional neural network (CNN)-based models, into the fullmesh, hubless adhoc VSATPlus<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">®</sup> platform, and examines the potential benefits of a more efficient and robust alternative to conventional algorithm-based channel coding schemes that do not leverage CNN models. The proposed CNN-based LDPC decoder, inspired by the belief propagation (BP)-CNN framework, refines noise estimation and error correction through iterative interactions between an LDPC decoder and a CNN module, effectively exploiting residual noise features inherent to satellite communication environments. Through CNN-based LDPC decoding, the system achieves enhanced error correction performance, reduced computational complexity, and improved adaptability to non-Gaussian noise scenarios, ensuring robust operation under multi-frequency time division multiple access (MF-TDMA) schemes. Furthermore, this work explores the practical deployment of the CNN-based LDPC decoder on Xilinx field-programmable gate array (FPGA) platforms using the Vitis artificial intelligence (AI) development framework. The hardware-aware design ensures resource-efficient implementation and real-time feasibility on the Zynq UltraScale+ multiprocessor system-on-chip (MPSoC). Simulation results further demonstrate the scalability and practical applicability of the proposed solution in real-world hubless full-mesh satellite networks by achieving up to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{4} \text{d B}$</tex> signal-to-noise ratio (SNR) gain over conventional ECC decoding at a bit error rate (BER) <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$=10^{-5}$</tex>. This enhancement suggests significantly improved transmission quality while reducing antenna size requirements for VSATPlus terminals. The findings validate the feasibility of deep learning-assisted LDPC decoding for next-generation hubless VSATPlus networks and pave the way for further research into adaptive neural-assisted error correction techniques.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.