Digital Pre-Distortion with Deep Reinforcement Learning for 5G Power Amplifiers

The relentless pursuit of higher data rates and improved energy efficiency in wireless communication systems forces power amplifiers (PAs) into non-linear operation, creating signal distortions that hinder quality and efficiency. Digital Pre-Distortion (DPD) tackles these distortions, but existing systems struggle with complexity, adaptability, and resource limitations. This paper presents a novel DPD solution, based on Deep Reinforcement Learning (DRL), called DRL-DPD, with the aim to significantly reduce the computational burden, enhance adaptation to dynamic environments, and minimize resource consumption compared to traditional approaches for both local and remote DPD configurations. Simulations and hardware experiments showcased the potential of the proposed method to surpass the limitations of current DPD techniques in both local and remote setups, paving the way for more efficient and effective communication systems, especially in the context of 5G and beyond technologies.