[Paper Review] Artificial Intelligence-aided Receiver for A CP-Free OFDM System: Design, Simulation, and Experimental Test

Jing Zhang, Chao-Kai Wen+2 more|arXiv (Cornell University)|Mar 12, 2019

Wireless Signal Modulation Classification30 references4 citations

TL;DR

This paper proposes an AI-aided receiver for CP-free OFDM systems using a model-driven deep learning approach, combining a least-squares-initialized channel estimation network (CE-NET) and an OAMP-based signal detection network (OAMP-NET). The receiver achieves superior BER performance—especially in high-order modulation and dynamic environments—demonstrated through simulations and real-world over-the-air (OTA) testing with lower complexity and greater robustness than existing methods.

ABSTRACT

Orthogonal frequency division multiplexing (OFDM), usually with sufficient cyclic prefix (CP), has been widely applied in various communication systems. The CP in OFDM consumes additional resource and reduces spectrum and energy efficiency. However, channel estimation and signal detection are very challenging for CP-free OFDM systems. In this paper, we propose a novel artificial intelligence (AI)-aided receiver (AI receiver) for a CP-free OFDM system. The AI receiver includes a channel estimation neural network (CE-NET) and a signal detection neural network based on orthogonal approximate message passing (OAMP), called OAMP-NET. The CE-NET is initialized by the least-square channel estimation algorithm and refined by a linear minimum mean-squared error neural network. The OAMP-NET is established by unfolding the iterative OAMP algorithm and adding several trainable parameters to improve the detection performance. We first investigate their performance under different channel models through extensive simulation and then establish a real transmission system using a 5G rapid prototyping system for an over-the-air (OTA) test. Based on our study, the AI receiver can estimate time-varying channels with a single training phase. It also has great robustness to various imperfections and has better performance than those competitive algorithms, especially for high-order modulation. The OTA test further verifies its feasibility to real environments and indicates its potential for future communications systems.

Motivation & Objective

• Address the challenge of channel estimation and signal detection in CP-free OFDM systems, where cyclic prefix removal reduces spectral efficiency but introduces inter-carrier and inter-block interference.
• Overcome the limitations of conventional methods like LS and ZF in time-varying and frequency-selective fading channels by leveraging AI for robustness and accuracy.
• Design a model-driven deep learning receiver that integrates physical-layer knowledge (LMMSE, OAMP) with trainable parameters to reduce training data needs and improve generalization.
• Validate the proposed AI receiver in both simulated and real-world over-the-air (OTA) environments, including dynamic scattering and mobility.
• Demonstrate the feasibility and superiority of the AI-aided receiver for future high-efficiency, low-latency wireless systems such as 6G and IoT.

Proposed method

• Design a channel estimation neural network (CE-NET) initialized with least-squares estimation and refined via a linear minimum mean-squared error (LMMSE) layer to improve accuracy with minimal data.
• Develop an OAMP-NET by unfolding the iterative orthogonal approximate message passing (OAMP) algorithm and introducing trainable parameters to enhance signal detection performance.
• Train the AI receiver offline using channel models such as the exponential power delay profile (EXP), representative of IEEE 802.11b, to simulate realistic multipath fading.
• Integrate both CE-NET and OAMP-NET into a unified end-to-end AI receiver architecture for joint channel estimation and signal detection in CP-free OFDM.
• Use a 5G rapid prototyping system to implement and test the AI receiver in real over-the-air (OTA) environments with controlled mobility and scattering.
• Compare performance against baseline LS+OFDM and other AI baselines (FC-DNN, ComNet) under identical SNR and pilot configurations in both static and dynamic scenarios.

Experimental results

Research questions

• RQ1 Can a model-driven AI receiver achieve accurate channel estimation and signal detection in CP-free OFDM systems with minimal training data and low complexity?
• RQ2 How does the proposed AI receiver perform in dynamic, real-world environments with moving scatterers compared to conventional and existing AI-based receivers?
• RQ3 To what extent does the OAMP-NET improve detection accuracy over zero-forcing-based methods like ComNet in high-order modulation and time-varying channels?
• RQ4 Does the AI receiver maintain robustness and low BER when trained on a single channel model (e.g., EXP) but tested in diverse real-world OTA conditions?
• RQ5 What is the performance gap between continuous and comb pilot configurations in the AI receiver, and how can it be mitigated?

Key findings

• The proposed AI receiver achieves a BER of $4.6 imes 10^{-3}$ in Scenario 2 (dynamic environment with moving people), significantly outperforming LS+OFDM ($1.9 imes 10^{-2}$), FC-DNN ($2.1 imes 10^{-2}$), and ComNet ($1.2 imes 10^{-2}$).
• In Scenario 1 (static environment), the AI receiver achieves BER of $3.7 imes 10^{-4}$, comparable to LS+OFDM ($3.9 imes 10^{-4}$) and slightly better than FC-DNN ($5.4 imes 10^{-4}$) and ComNet ($7.8 imes 10^{-4}$).
• The OAMP-NET detector demonstrates higher precision than ZF-based ComNet, especially in high-order modulation and complex fading environments.
• The AI receiver shows strong robustness to channel variations, modulation mode changes, and pilot structure (comb vs. continuous), maintaining low BER across diverse conditions.
• The OTA test confirms the feasibility of the AI receiver in real-world environments, validating its performance advantage over existing methods in practical, non-ideal conditions.
• The receiver achieves lower complexity than traditional model-based receivers while approaching the theoretical performance limit, particularly in high-SNR and high-mobility scenarios.