Generative AI-enabled Wireless Communications for Robust Low-Altitude Economy Networking

Changyuan Zhao, Jiacheng Wang, Ruichen Zhang, Dusit Niyato, Geng Sun, Hongyang Du · 2025

Read Generative AI-enabled Wireless Communications for Robust Low-Altitude Economy Networking by Changyuan Zhao, Jiacheng Wang, Ruichen Zhang, Dusit Niyato, Geng Sun, Hongyang Du (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-18118.txt.

Introduction

Changyuan Zhao and co-authors surveyed how generative AI could make low flying drone networks tougher, in a 2025 paper. These networks deliver parcels, food, and medical goods, but their moving parts and open air links leave them open to attack.

The authors first set out what toughness means for such networks and review measures of service quality at the radio level. They then survey current generative methods used to harden networks. Their own proposal uses a diffusion style model with a team of expert sub networks to steer the radio beams. In a case study on secure beam steering under uncertainty, it beat four learning baselines by more than fifteen percent on the worst case safe rate.

Catalog record

CREATORChangyuan Zhao, Jiacheng Wang, Ruichen Zhang, Dusit Niyato, Geng Sun, Hongyang Du
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-18118

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Generative AI-enabled Wireless Communications for Robust Low-Altitude Economy Networking Thanks: C. Zhao is with the College of Computing and Data Science, Nanyang Technological University, Singapore, and CNRS@CREATE, 1 Create Way, 08-01 Create Tower, Singapore 138602 (e-mail: [email protected]). Thanks: J. Wang, R. Zhang, and D. Niyato are with the School of Computer Science and Engineering, Nanyang Technological University, Singapore (e-mail: [email protected], [email protected], [email protected]). Thanks: G. Sun is with College of Computer Science and Technology, Jilin University, China 130012, (e-mail: [email protected]). Thanks: H. Du is with the Department of Electrical and Electronic Engineering, University of Hong Kong, Hong Kong (e-mail: [email protected]). Thanks: D. I. Kim is with the Department of Electrical and Computer Engineering, Sungkyunkwan University, Suwon 16419, South Korea (e-mail:[email protected]). Thanks: A. Jamalipour is with the School of Electrical and Computer Engineering, University of Sydney, Australia (e-mail: [email protected]). Changyuan Zhao, Jiacheng Wang, Ruichen Zhang, Dusit Niyato, Geng Sun, Hongyang Du Abstract Low-Altitude Economy Networks (LAENets) have emerged as significant enablers of social activities, offering low-altitude services such as the transportation of packages, groceries, and medical supplies. Unlike traditional terrestrial networks, LAENets are characterized by control mechanisms and ever-changing operational factors, which make them more complex and susceptible to vulnerabilities. As applications of LAENet continue to expand, robustness of these systems becomes crucial. In this paper, we investigate a novel application of Generative Artificial Intelligence (GenAI) to improve the robustness of LAENets. We conduct a systematic analysis of robustness requirements for LAENets, complemented by a comprehensive review of robust Quality of Service (QoS) metrics from the wireless physical layer perspective. We then investigate existing GenAI-enabled approaches for robustness enhancement. This leads to our proposal of a novel diffusion-based optimization framework with a Mixture of Expert (MoE)-transformer actor network. In the robust beamforming case study, the proposed framework demonstrates its effectiveness by optimizing beamforming under uncertainties, achieving a more than 44% increase in the worst-case achievable secrecy rate. These findings highlight the significant potential of GenAI in strengthening LAENet robustness. I Introduction Recently, advancements in wireless communications and manufacturing technologies have propelled the development of drones and other aerial devices, greatly expanding their applications in low-altitude airspace. Enabled by these technologies, the Low-Altitude Economy (LAE) has emerged as a significant sector focusing on social opportunities within low-altitude airspace (defined as elevations up to 1,000 meters). The LAE aims to harness low-flying equipment for various social activities, including air traffic management, transportation, and surveillance, to organize the Low-Altitude Economy Networks (LAENet). These networks are essential as they provide critical services such as communication, computation, storage, and sensing [ 1 ] . However, the unique low-altitude environment and the absence of unified management regulations have introduced several uncertainties to the design of LAENet. These uncertainties, including variable weather conditions, air traffic, user demand, and emergency situations, significantly heighten the requirement for LAENet devices to complete tasks in dynamic environments [ 1 ] . Therefore, robustness has been recognized as a Key Performance Indicator (KPI) that goes beyond efficiency (speed) and effectiveness (accuracy) for LAENet. Robustness refers to the ability to maintain the qualities of service (QoS) of LAENets despite the influence of uncertainties [ 2 ] . This recognition guides the LAENet’s evolution, enabling innovation and commercial activity. For example, the tech giant Uber’s drone delivery utilizes advanced artificial intelligence algorithms, GPS systems, and sensors to avoid obstacles under different weather conditions, ensuring maximum delivery safety. However, due to the complexity of the airspace environment, the high degrees of freedom, and the strong dynamics of aerial devices, building a robust wireless network platform for LAE applications is challenging. Enhancing the robustness of the wireless physical layer requires addressing uncertainty as a critical issue. Zero-forcing techniques are applicable for accurate signal detection [ 3 ] . Deep learning-based joint source-channel coding can provide a reliable method for data transmission [ 4 ] . However, both numerical algorithms and traditional learning methods face several challenges, including heavy dependence on training data, and susceptibility to unknown environments, which hinder their effectiveness in handling uncertainty. Generative artificial intelligence (GenAI) is an emerging technology that works to analyze data distribution. GenAI has demonstrated strong capabilities in enhancing wireless communication, including physical and network layers, through cross-dimensional data processing and generation [ 5 ] . Based on the achievements in wireless communication, there are several potential advantages of using GenAI to eliminate uncertainty and construct robust LAENets. GenAI can generate additional signal data, which is difficult to obtain in extreme scenarios, to handle diverse interference patterns. Additionally, GenAI can generate clean data, thereby removing the noise uncertainty in signals. Moreover, GenAI can learn patterns from channel transmission data to realistically simulate fading effects and provide precise uncertainty estimates. GenAI models with transformer architectures, such as Large Language Models (LLMs) and Deepseek [ 6 ] , effectively capture long-range dependencies and contextual information. They are robust against noisy inputs and distribution shifts, making them useful for semantic communication, adaptive coding, and efficient spectrum utilization in wireless networks [ 7 ] . Mixture of Experts (MoE) architectures enhance computational efficiency by activating relevant expert pathways, improving scalability and reducing errors by adapting to changing channel conditions [ 8 ] . Motivated by these analyses, this paper aims to investigate the incorporation of GenAI into the development of robust LAENets. The main contributions are summarized as follows: • We provide an overview of LAENets and their applications, offering a detailed analysis of the robustness within LAENet. Furthermore, we introduce various GenAI models and their applications in wireless communications toward LAENets. • From the perspective of the physical layer, we explore how GenAI enhances the robustness of wireless communication. Moreover, we present several robustness metrics and discuss the benefits of implementing GenAI models to support the construction of LAENets. • We propose an MoE-transformer actor network integrated with the diffusion-based optimization framework in LAENets. Through a case study, we illustrate how GenAI can facilitate the development of robust beamforming for LAENets. II The overall of Robust LAENet and GenAI In this section, we first introduce the LAENet and its related applications. Further, we discuss the robust issues and challenges in the LAENets. Fig. 1 : An overall of GenAI-enabled robust LAENet. Part A illustrates the architecture of LAENet from transportation, communication, and sensing. Part B introduces three aspects of robustness in LAENets. Part C presents a comparison between GenAI and deterministic AI. Part D demonstrates important GAI-enabled applications, including channel modeling, channel estimation, and beam alignment. II-A Conceptions and Applications of LAENet The LAE relies on the cooperation of edge equipment and low-altitude communication networks to facilitate activities, which covers three primary aspects (shown in Fig. 1 Part A ): • Low-Altitude Transportation : One of the most prominent applications of low-altitude transportation is drone-based delivery. Companies including Amazon, Walmart, and Wing are pioneering the use of drones to transport packages, groceries, and medical supplies. • Low-Altitude Communication : LAENets can address the limitations of traditional communication systems in remote and rural areas. For instance, Telia Company has deployed drones equipped with portable mobile base stations to enable remote control of forestry machinery over 5G networks. • Low-Altitude Sensing : Drones are invaluable tools for monitoring environmental changes. Drone enables high-resolution mapping and data collection

Read the complete text