A Review on ANN-Based Behavioral Modeling Techniques for High-Efficiency Power Amplifiers

This paper presents a comprehensive review of artificial neural network (ANN)-based behavioral modeling techniques for high-efficiency power amplifiers (PAs). As wireless systems demand greater linearity and efficiency, accurate modeling of PA nonlinearity and memory effects becomes essential. We systematically explore fundamental and advanced ANN architectures including memoryless networks, and memory networks such as temporal encoding, recurrent models, knowledge infused networks, attention mechanisms, and hybrid approaches, and evaluate their capabilities in capturing complex PA dynamics. The study further investigates optimization algorithms for enhancing ANN model performance and their integration into digital predistortion (DPD) systems for linearization. A dedicated section discusses embedded deployment considerations, highlighting trade-offs between accuracy, complexity, and real time applicability on platforms such as field-programmable gate arrays (FPGAs) and digital signal processors (DSPs). Finally, we provide architectural selection guidelines and identify future research directions, including lightweight model design, multi band modeling, and automated ANN-driven PA optimization. This review demonstrates the practical utility of ANN-based techniques for advancing PA design in next-generation communication systems.