Abstract
The increasing complexity of wireless environments, driven by user mobility and dynamic obstructions, poses significant challenges to maintaining Line-of-Sight (LoS) connectivity. Mobile base stations (gNBs) offer a promising solution by physically relocating to restore or sustain LoS. This paper explores how reinforcement learning (RL) can be applied to gNB mobility control within vision-aided network systems. As part of the CONVERGE project, we present the CONVERGE Chamber Simulator (CC-SIM), a 3D environment for developing, training, and testing gNB mobility control algorithms. CC-SIM models user and obstacle mobility, visual occlusion, and Radio Frequency (RF) propagation while supporting both offline reinforcement learning and real-time validation through integration with OpenAirInterface (OAI). Leveraging CC-SIM, we trained a Deep Q-Network (DQN) agent that proactively repositions gNBs under dynamic conditions. Across three representative use cases, the agent reduced LoS blockage by up to 42% compared to static deployments, highlighting the potential of RL-driven mobility control and positioning CC-SIM as a practical platform for advancing adaptive, next-generation wireless networks.

Abstract
Unmanned Aerial Vehicles (UAVs) are a promising solution for next-generation wireless networks due to their mobility, rapid deployment, and ability to provide Line-of-Sight (LoS) connectivity. However, deploying multiple UAVs in realt-ime to meet dynamic, non-uniform traffic demands remains a significant challenge, especially when aiming to optimize network throughput and resource utilization. In this paper, we propose the Efficient Multi-UAV Traffic-Aware Deployment (EMTAD) algorithm, a scalable algorithm that jointly minimizes UAV count and optimizes 3D positioning based on real-time user distribution and traffic demand. In contrast to prior works that assume static user patterns or fixed UAV counts, EMTAD dynamically adapts UAV deployment to maximize spectral efficiency and satisfy user-specific Quality of Service (QoS) requirements. Simulation results demonstrate that EMTAD reduces the number of UAVs required and achieves superior aggregate throughput compared to baseline approaches.

Abstract
High-frequency radio networks, including those operating in the millimeter-wave bands, are sensible to Line-of-Sight (LoS) obstructions. Computer Vision (CV) algorithms can be leveraged to improve network performance by processing and interpreting visual data, enabling obstacle avoidance and ensuring LoS signal propagation. We propose a vision-aided Radio Access Network (RAN) based on the O-RAN architecture and capable of perceiving the surrounding environment. The vision-aided RAN consists of a gNodeB (gNB) equipped with a video camera that employs CV techniques to extract critical environmental information. An xApp is used to collect and process metrics from the RAN and receive data from a Vision Module (VM). This enhances the RAN's ability to perceive its surroundings, leading to better connectivity in challenging environments.

Abstract
Unmanned Aerial Vehicles (UAVs) increasingly enhance the Quality of Service (QoS) in wireless networks due to their flexibility and cost-effectiveness. However, optimizing UAV placement in dynamic, obstacle-prone environments remains a significant research challenge due to their complexity. Reinforcement Learning (RL) offers adaptability and robustness in such environments, proving effective for UAV positioning optimization. This paper introduces RLpos-3, a novel framework that integrates standard RL techniques and simulation libraries with Network Simulator 3 (ns-3) to facilitate the development and evaluation of UAV positioning algorithms. RLpos-3 serves as a supplementary tool for researchers, enabling the implementation, analysis, and benchmarking of UAV positioning strategies across diverse environmental conditions while meeting user traffic demands. To validate its effectiveness, we present use cases demonstrating RLpos-3's performance in optimizing UAV placement under realistic conditions, such as urban and obstacle-rich environments.

Abstract
Integrated Access and Backhaul (IAB) in cellular networks combines access and backhaul within a wireless infrastructure reducing reliance on fibre-based backhaul. This enables flexible and more cost-effective network expansion, especially in hard-to-reach areas. Positioning a mobile IAB node (MIAB) in a seaport environment, in order to ensure on-demand, resilient wireless connectivity, presents unique challenges due to the high density of User Equipments (UEs) and potential shadowing effects caused by obstacles. This paper addresses the problem of positioning MIABs within areas containing UEs, fixed IAB donors (FIABs), and obstacles. Our approach considers user associations and different types of scheduling, ensuring MIABs and FIABs meet the capacity requirements of a special team of served UEs, while not exceeding backhaul capacity. With a Genetic Algorithm solver, we achieve capacity improvement gains, by up to 200% for the 90th percentile, particularly during emergency capacity demands.

Abstract