Precoding is a technique employed to enhance transmission rates in various communication systems, including massive multiple-input multiple-output (MIMO) and optical wireless communication (OWC). In this study, we focus on network massive MIMO OWC (NM-MIMO-OWC) systems and investigate the precoder design to enhance the sum rate and improve the system performance. We present the network’s massive MIMO OWC framework. By utilizing this framework, we are able to calculate the achievable sum rate. Subsequently, we consider the precoding design for maximizing the sum rate while adhering to the total power constraint. To solve this optimization problem, we provide a necessary condition of the optimal solution based on the Karush–Kuhn–Tucker (KKT) conditions, and propose a low-complexity algorithm to further enhance the efficiency of the proposed precoding technique. The numerical results demonstrate that the proposed precoder design significantly improves the transmission rate and effectively maximizes the sum rate.

Multiple-input multiple-output (MIMO) optical wireless communications (OWC) is a key technology to meet the growing demand for high data rates and reliable connectivity in sixth-generation (6G) wireless communications. This paper proposes a novel precoding framework for a user-centric network massive MIMO OWC system, using both the iterative weighted minimum mean squared error (WMMSE) and the deep learning-based Transformer model. The objective of the proposed precoding technique is to maximize the achievable sum rate under a total power constraint. In user-centric network MIMO OWC, each user is dynamically operated by multiple distributed access points (APs), thereby improving coverage and reducing inter-user interference. The design of optimal precoding in large-scale and condensed networks remains a crucial challenge due to the large number of transmitters and receivers. Traditional optimization-based precoding algorithms can achieve near-optimal performance, but, their computational complexity is high due to large-scale matrix operations. To tackle this problem, we first propose an iterative WMMSE precoding algorithm that achieves the sub-optimal solution and serves as a benchmark. Following this, we present a Transformer-based precoding approach to achieve near sub-optimal performance. The Transformer model is trained through the available channel information to predict precoding matrices efficiently with low computational complexity. The numerical results show that the proposed Transformer network model achieves better performance with low computation complexity compared to the existing conventional precoding techniques.

Successive cancelation list (SCL) decoding for polar codes has the ability to achieve Shannon capacity over the binary discrete memoryless channel. However, a large list size is required to achieve the Shannon capacity limit. Consequently, this article proposes a novel iterative method for the CA-SCL decoder based on frozen bits to improve the error-correction performance of CA-SCL decoding for polar codes with smaller list sizes. The proposed technique determines the vulnerable channel log-likelihood ratios (LLRs) for frozen bits that degrade the CA-SCL decoding error correction performance and corrects them for the next CA-SCL decoding attempt. We analyze the behavior of wrongly estimated LLRs of frozen bits during CA-SCL decoding and design an attenuation factor set using a greedy algorithm. The proposed technique employs iterative CA-SCL decoding and revises the wrongly estimated frozen bit LLRs using the attenuation factor set during the decoding process. The simulation results show that the proposed technique achieves a block error rate (BLER) performance gain of up to 0.25 compared with the CA-SCL decoder, with a minor increase in computational complexity. Moreover, the attenuation factor set can be adjusted arbitrarily to balance performance and complexity.