Design and Optimization of Machine Learning-Enhanced Forward Error Correction Codes for Improving Data Integrity in Global Navigation Satellite Networks

Abstract

A robust error correction method is essential to address the growing demand for accurate data transmission in global Navigation satellite network reliably. Turbo Code Decoder (TCD) and Viterbi Decoder (VD) convolutional coding systems are in wide use due to their capability in improving the data integrity. These methods have limitation to handle high bit error rate and to adapt to continuous changing signal condition found in urban environment. This work presents machine learning based improved Forward Error Correction (ML-FEC) System. This system combines Convolutional Neural Network with conventional Error rectification process. The proposed ML-FEC improves reduction in complexity of computation, Error detection and correction capabilities. This is achieved by incorporating a novel approach of combining Low Density Parity Check algorithm with machine learning model that can adjust to transmission conditions in real time. The error correction efficiency has been improved by adopting new methods like reinforcement learning based parameters and adaptive learning rate adjustments. The proposed system has achieved 0.20% reduction in Bit Error Rate and an improvement of 0.15% in throughput of complete data in comparison to conventional systems. In addition, the load of computation is decreased by 0.25% which addresses the main drawbacks of increased complexity in existing systems. From all these observations, ML-FEC offers to be a viable solution for present systems, by offering increases performance and reliability in complex transmission environments.