The evolution of network technologies has witnessed a paradigm shift toward open and intelligent networks, with the Open Radio Access Network (O-RAN) architecture emerging as a promising solution. O-RAN introduces disaggregation and virtualization, enabling network operators to deploy multi-vendor and interoperable solutions. However, managing and automating the complex O-RAN ecosystem presents numerous challenges. To address this, machine learning (ML) techniques have gained considerable attention in recent years, offering promising avenues for network automation in O-RAN. This paper presents a comprehensive survey of the current research efforts on network automation usingML in O-RAN.We begin by providing an overview of the O-RAN architecture and its key components, highlighting the need for automation. Subsequently, we delve into O-RAN support forML techniques. The survey then explores challenges in network automation usingML within the O-RAN environment, followed by the existing research studies discussing application of ML algorithms and frameworks for network automation in O-RAN. The survey further discusses the research opportunities by identifying important aspects whereML techniques can benefit.

Powered by deep learning (DL), autoencoders (AE) end-to-end (E2E) communication systems have been developed to merge all physical layer blocks in traditional communication systems and have achieved great success. In this paper, a new probabilistic model, based on the variational autoencoders (VAE), is proposed for short packet wireless communication systems. Using this new approach, the information messages are represented by the so called packet hot vectors (PHV) which are inferred by the VAE latent random variables (LRVs). Then only LRVs’ parameters can be transmitted through the physical wireless channel. This results in a significant improvement in spectral efficiency when compared with the pure AE approach where a longer hot vectors are to be transmitted. Specific VAE models have been developed for both binary (BPSK) as well as Quadrature phase shift keying (QPSK) systems. Simulation and numerical results are given to demonstrate the performance of the proposed method in different real scenarios, including Rayleigh and Rician fading channels with Shadowing and Doppler effects. Our simulation and numerical results show that the new proposed VAE with a DL classifier can provide an improved symbol error rate (SER) performance than both the baseline AE and the classical Hamming code with hard decision decoding. Furthermore, as far as the spectral efficiency of the proposed method is concerned, we show that using two channels in the proposed VAE performance exceeds the 7 channels baseline AE.

Open Radio Access Networks (O-RANs) have revolutionized the telecom ecosystem by bringing intelligence into disaggregated RAN and implementing functionalities as Virtual Network Functions (VNF) through open interfaces. However, dynamic traffic conditions in real-life O-RAN environments may require necessary VNF reconfigurations during run-time, which introduce additional overhead costs and traffic instability. To address this challenge, we propose a multi-objective optimization problem that minimizes VNF computational costs and overhead of periodical reconfigurations simultaneously. Our solution uses constrained combinatorial optimization with deep reinforcement learning, where an agent minimizes a penalized cost function calculated by the proposed optimization problem. The evaluation of our proposed solution demonstrates significant enhancements, achieving up to 76% reduction in VNF reconfiguration overhead, with only a slight increase of up to 23% in computational costs. In addition, when compared to the most robust O-RAN system that doesn’t require VNF reconfigurations, which is Centralized RAN (C-RAN), our solution offers up to 76% savings in bandwidth while showing up to 27% overprovisioning of CPU.

This article proposes a downlink/uplink decoupled (DUDe) access scheme for cellular-enabled unmanned aerial vehicle (UAV) communication systems. To minimize interference, the proposed scheme separates the control and data links of UAVs, as well as the uplinks (ULs) and downlinks (DLs) of ground users (GUEs), onto different serving base stations and operating frequencies. Since power availability is a major constraint in UAV communications, two power allocation schemes based on $Q$ -learning (QL) and deep $Q$ -learning (DQL) are proposed to optimize the communication energy efficiency (EE) of this DUDe network. To quantify the improvements achieved, the proposed schemes are compared with the fractional power control (FPC) scheme used in 4G and 5G networks, as well as a convex optimization-based optimal power allocation scheme. The results demonstrate that the proposed DUDe scheme can achieve up to several times higher sum rates and EE in the UL direction than its coupled counterparts. Moreover, it is shown that the EE performance of the QL and DQL power allocation schemes approach the optimal performance and surpass the conventional FPC scheme by 80%–100% in the UHF band, and by 160%–170% in the mmWave band.