An Optimized Adaptive Bayesian Algorithm for Mitigating EMI-Induced Errors in Dynamic Electromagnetic Environments

• Published in: IEEE transactions on electromagnetic compatibility Vol. 66; no. 6; pp. 2085 - 2094
• Main Authors: Gonzalez-Atienza, Miriam, Vanoost, Dries, Verbeke, Mathias, Pissoort, Davy
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.12.2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Adaptive algorithm; Adaptive algorithms; Adaptive learning; Adaptive systems; Algorithms; Bayes methods; Bayesian analysis; Bayesian learning; Computational modeling; electromagnetic environment; Electromagnetic interference; electromagnetic interference (EMI); Electromagnetics; Electronic systems; Encoding; Machine learning; Performance enhancement; Performance evaluation; Posterior probability; Supervised learning; Symbols; Trajectory; Uncertainty
• ISSN: 0018-9375; 1558-187X
• DOI: 10.1109/TEMC.2024.3482565

Robust and reliable communication is a necessity for maintaining the integrity of electronic systems. However, this can be severely impacted by electromagnetic interference (EMI) and the increasing complexity of electromagnetic environments, which introduces new sources of uncertainty. Fortunately, adaptive learning strategies provide promising solutions to overcome these obstacles. In this study, we have evaluated the performance of an adaptive Bayesian decision algorithm that can effectively mitigate the impact of EMI in triple modular redundant channels subjected to multiple single-frequency disturbances. The proposed strategy for hyperparameter tuning allows for optimization of data accuracy and availability. Compared to conventional supervised machine learning techniques, the inherent adaptive capabilities of the Bayesian approach demonstrate superior performance by dynamically adjusting to changes in the received symbol distributions, particularly under conditions of high symbol error rates. This adaptability proved crucial in achieving high classification accuracy (92.59%). The proposed adaptive Bayesian algorithm stands out for its robustness and enhanced performance, presenting a promising solution for its application to dynamic electromagnetic environments characterized by high variability and uncertainty.