Despite significant advancements in deep learning-based CSI compression, some key limitations remain unaddressed. Current approaches predominantly treat CSI compression as a source-coding problem, thereby neglecting transmission errors. Conventional separate source and channel coding suffers from the cliff effect, leading to significant deterioration in reconstruction performance under challenging channel conditions. While existing autoencoder-based compression schemes can be readily extended to support joint source-channel coding, they struggle to capture complex channel distributions and exhibit poor scalability with increasing parameter count. To overcome these inherent limitations of autoencoder-based approaches, we propose Residual-Diffusion Joint Source-Channel Coding (RD-JSCC), a novel framework that integrates a lightweight autoencoder with a residual diffusion module to iteratively refine CSI reconstruction. Our flexible decoding strategy balances computational efficiency and performance by dynamically switching between low-complexity autoencoder decoding and sophisticated diffusion-based refinement based on channel conditions. Comprehensive simulations demonstrate that RD-JSCC significantly outperforms existing autoencoder-based approaches in challenging wireless environments. Furthermore, RD-JSCC offers several practical features, including a low-latency 2-step diffusion during inference, support for multiple compression rates with a single model, robustness to fixed-bit quantization, and adaptability to imperfect channel estimation.

We propose a machine learning (ML) based end-to-end framework for pilotless communications that consists of two key components. The first component is an asymmetric modulation constellation that enables pilotless communications under channel impairments. The second component is a neural network (NN) receiver featuring an architecture that has a core of several serially-connected ResNet-like blocks. The transmitter only sends data symbols (without any pilots), and the NN receiver enables pilotless communications by using the received data symbols from the asymmetric constellation to perform implicit channel estimation/compensation and generate log-likelihood ratios (LLRs) for the bits comprising the data symbols. The combination of the asymmetric modulation constellation and the NN receiver achieves similar or superior performance to a traditional zero-forcing (ZF) receiver that relies on pilot symbols for channel estimation for 64-ary and 256-ary modulations for channels with limited time and frequency selectivity.

This paper presents a puncturing pattern design for polar codes in hybrid automatic repeat request (HARQ) systems. We divide the encoder output bit sequence into sub-blocks. The sub-block indices are then permuted to generate a set of equivalent puncturing patterns (EPPs), each of which guarantees high reliability for message bits. A puncturing pattern for each HARQ transmission is selected from the generated set of EPPs. A Gaussian-approximation based method is proposed to optimize the selection, ensuring that additional redundancy bits are sent in HARQ transmissions. The proposed design has low complexity due to the limited number of sub-blocks and can be implemented offline to generate a lookup table of puncturing patterns. Up to 0.5 dB improvements in the error-correcting performance are obtained compared with state-of-the-art techniques.

We propose machine learning (ML) based optimization methods for modulation and digital predistortion (DPD), that overcome the signal distortion due to power amplifier (PA) non-linearity and memory effects. The proposed methods generate and exploit an adjusted modulation constellation to compensate for a given PA non-linearity, whereas conventional and widely-employed methods mainly rely on DPD to that end. This potentially removes the need for DPD if memory effects are not detrimental and enables the transmitter to operate close to the power saturation region, increasing the PA power efficiency. The AI/ML framework to learn an adjusted constellation is trained to produce a target square quadrature amplitude modulation (QAM) signal at the PA output. We also present a DPD learning architecture for the adjusted constellations. The proposed methods outperform square 16-ary/64-ary QAMs with DPD by more than 1 dB and are within 0.2~10.3 dB of the theoretical performance at a symbol error rate of 0.01, when the PA operates in its saturation region.

Tailoring polar code construction for decoding algorithms beyond successive cancellation has remained a topic of significant interest in the field. However, despite the inherent nested structure of polar codes, the use of sequence models in polar code construction is understudied. In this work, we propose using a sequence modeling framework to iteratively construct a polar code for any given length and rate under various channel conditions. Simulations show that polar codes designed via sequential modeling using transformers outperform both 5G-NR sequence and Density Evolution based approaches for both AWGN and Rayleigh fading channels.

A two-user interference channel is a canonical model for multiple one-to-one communications, where two transmitters wish to communicate with their receivers via a shared medium, examples of which include pairs of base stations and handsets near the cell boundary that suffer from interference. Practical codes and the fundamental limit of communications are unknown for interference channels as mathematical analysis becomes intractable. Hence, simple heuristic coding schemes are used in practice to mitigate interference, e.g., time division, treating interference as noise, and successive interference cancellation. These schemes are nearly optimal for extreme cases: when interference is strong or weak. However, there is no optimality guarantee for channels with moderate interference. Here we combine deep learning and network information theory to overcome the limitation on the tractability of analysis and construct finite-blocklength coding schemes for channels with various interference levels. We show that carefully designed and trained neural codes using network information theoretic insight can achieve several orders of reliability improvement for channels with moderate interference. Furthermore, we present the interpretation of the learned codes based on the codeword distance and the Centered Kernel Alignment (CKA) analysis.

We propose machine learning (ML) based optimization methods and new high order modulations for reliable and high-capacity communications. The widely adopted square quadrature amplitude modulations (QAM) fundamentally exhibit a shaping loss of up to 1.53 dB to the Shannon capacity bound. The proposed modulations obtained through the ML based optimization outperform the square QAMs and other state of-the-art ones by about 1.2 dB and 0.3 dB, respectively, for 1024-ary modulation with LDPC coding. We construct the neural network architecture and training methods to reflect the desired properties of well-performing modulations. This significantly helps in the training convergence of the ML models to a desired optimal state and leads to the modulation constellation and bit to-symbol mapping that reduces the shaping loss to the Shannon capacity bound to a large extent. Moreover, the ML methods enable the development of new optimal modulations for a wide range of target SNR and modulation orders.

A critical aspect of reliable communication involves the design of codes that allow transmissions to be robustly and computationally efficiently decoded under noisy conditions. Advances in the design of reliable codes have been driven by coding theory and have been sporadic. Recently, it is shown that channel codes that are comparable to modern codes can be learned solely via deep learning. In particular, Turbo Autoencoder (TurboAE), introduced by Jiang et al., is shown to achieve the reliability of Turbo codes for Additive White Gaussian Noise channels.In this paper, we focus on applying the idea of TurboAE to various practical channels, such as fading channels and chirp noise channels. We introduce TurboAE-TI, a novel neural architecture that combines TurboAE with a trainable interleaver design. We develop a carefully-designed training procedure and a novel interleaver penalty function that are crucial in learning the interleaver and TurboAE jointly. We demonstrate that TurboAE-TI outperforms TurboAE and LTE Turbo codes for several channels of interest. We also provide interpretation analysis to better understand TurboAE-TI.

In multiple-input multiple-output (MIMO) systems, the high-resolution channel information (CSI) is required at the base station (BS) to ensure optimal performance, especially in the case of multi-user MIMO (MU-MIMO) systems. In the absence of channel reciprocity in frequency division duplex (FDD) systems, the user needs to send the CSI to the BS. Often the large overhead associated with this CSI feedback in FDD systems becomes the bottleneck in improving the system performance. In this paper, we propose an AI-based CSI feedback based on an auto-encoder architecture that encodes the CSI at UE into a low-dimensional latent space and decodes it back at the BS by effectively reducing the feedback overhead while minimizing the loss during recovery. Our simulation results show that the AI-based proposed architecture outperforms the state-of-the-art high-resolution linear combination codebook using the DFT basis adopted in the 5G New Radio (NR) system.

As the pace of global 5G network deployments accelerates, the telecommunications industry strives to move from a conventional vertical stack, closed, hardware-based ecosystem to an open, interoperable, modular, cloud-based ecosystem that leverages software-based implementations of various network entities – including core network functions and base stations. Such an open ecosystem could facilitate the implementation of AI techniques and could be a critical advancement in realizing the vision of “zero-touch” wireless networks that support end-to-end automation– stretching from the core network to end-user devices. In this paper, we provide an overview of ongoing discussions along these lines in the context of the O-RAN Alliance along with a comprehensive discussion on various challenges. We then briefly present our vision of the evolution of intelligent networks beyond current 4G/5G networks.

High-resolution Type II CSI feedback is a key differentiator between 4G (LTE) and 5G (NR). The large feedback overhead of Type II CSI reporting, however, is a bottleneck for UE implementations in real deployments. In this paper, a space-frequency compression based Type II CSI overhead reduction has been proposed. The system-level simulation results are provided to show that the performance similar to Type II CSI is achievable with significantly reduced CSI overhead.

Current trends in spectrum regulation show that more and more unlicensed and shared spectrum bands are poised to be opened up for mobile communication. However, the question remains how to best utilize this spectrum and build efficient networks, and if the time has come for newer approaches to be considered for the next generation system. In this work, we propose a coordinated shared spectrum framework that can be considered for next generation cellular standardization. In designing the framework, we aim to improve on the current unlicensed access schemes toward increasing spectral efficiency in highly- dense networks. To this end, we demonstrate that with the proposed framework both throughput and access delay can be significantly improved over the state-of-the-art LAA system. We also show that large statistical multiplexing gains are possible through dynamic sharing instead of static, hard splitting of shared spectrum, as in the current CBRS system.

The LTE-Advanced (LTE-A) system is currently under development to allow for significantly higher spectral efficiency and data throughput than LTE systems. In a wireless system based on orthogonal frequency division multiplexing (OFDM) with frequency reuse factor one such as LTE, the achievable cell spectral efficiency is often limited by the inter-cell interference or coverage shortage of base stations. Hence in LTE-A, coordinated multi-point (CoMP) transmission/reception (a.k.a. multi-cell MIMO or base station cooperation) and relaying technologies are being introduced to clear these major performance hurdles. In this paper, overall picture of cooperative communication technologies being discussed in LTE-A systems including CoMP and relaying is presented, together with considerations on system design.

A novel frequency-hop spread-spectrum multiple-access network employing M-ary orthogonal Walsh sequence keying with noncoherent demodulation is proposed. The transmitted Walsh sequence is overlaid by a user-specific pseudonoise sequence to reduce the effect of multiple-access hits. Two Gaussian approximations for the multiple-access interference from both the dehopped slot and its neighboring slots are developed and are used to analyze the performance of the proposed network for synchronous and asynchronous hopping under nonfading and Rayleigh fading channels. The effect of imperfect hop timing synchronization at the receiver is also analyzed. It is shown that the proposed network offers significantly improved network throughput compared to networks based on traditional M-ary frequency-shift keying modulation.