Enhanced Channel Estimation for RIS-assisted OTFS Systems by Introducing ELM Network

Abstract In high-mobility communication scenarios, leveraging reconfigurable intelligent surfaces (RIS) to assist orthogonal time frequency space (OTFS) systems proves advantageous. However, the integration of RIS into OTFS systems has added complexity to the channel estimation (CE). Utilizing the benefits of machine learning (ML) to address such intricate issues holds the potential to reduce the complexity of CE. Yet, there is a lack of investigations of ML-based CE in RIS-assisted OTFS systems, leaving significant blanks and posing challenges for intelligent applications. Additionally, ML-based CE still faces numerous difficulties, such as intricate parameter tuning and long training time. Inspired by the inherent advantages of the single-hidden layer feed-forward network structure, we introduce extreme learning machine (ELM) into RIS-assisted OTFS systems to improve their CE accuracy. In this method, we integrate a threshold-based approach to extract initial feature, aiming to remedy the inherent limitations of ELM network, such as inadequate network parameters compared to the deep learning network. This initial feature is then leveraged to enhance ELM learning ability, leading to improved CE accuracy. By using the classic message passing algorithm to detect data symbols, simulation results demonstrate that the proposed method efficiently improves the symbol detection (SD) performance of RIS-assisted OTFS systems. Furthermore, the SD performance is robust against the impacts of the parameter variations.