In remote control systems, transmitting large data volumes (e.g., images, video frames) from wireless sensors to remote controllers is challenging when uplink capacity is limited (e.g., RedCap devices or massive wireless sensor networks). Furthermore, controllers often need only information-rich representations of the original data. To address this, we propose a semantic-driven predictive control combined with a channel-aware scheduling to enhance control performance for multiple devices under limited network capacity. At its core, the proposed framework, coined Time-Series Joint Embedding Predictive Architecture (TS-JEPA), encodes high-dimensional sensory data into low-dimensional semantic embeddings at the sensor, reducing communication overhead. Furthermore, TS-JEPA enables predictive inference by predicting future embeddings from current ones and predicted commands, which are directly used by a semantic actor model to compute control commands within the embedding space, eliminating the need to reconstruct raw data. To further enhance reliability and communication efficiency, a channel-aware scheduling is integrated to dynamically prioritize device transmissions based on channel conditions and age of information (AoI). Extensive simulations on inverted cart-pole systems demonstrate that the proposed framework achieves a 98.95% reduction in communication cost, a normalized prediction error of 0.004, and 74.48% control accuracy, outperforming conventional control baselines. Furthermore, the proposed framework maintains robust control performance on 15-step prediction horizons and supports up to 16× more devices than round-robin and opportunistic scheduling schemes with conventional control baselines under similar network conditions.

Integrating Large AI Models (LAMs) into 6G mobile networks is a key enabler of the AI-Native Air Interface (AI-AI), where protocol intelligence must scale beyond handcrafted logic. This paper presents, to our knowledge, the first standards-compliant emulation of the Radio Resource Control (RRC) layer using a decoder-only LAM (LLAMA-class) fine-tuned with Low-Rank Adaptation (LoRA) on a multi-vendor corpus of real-world traces spanning both 5G and 4G systems. We treat RRC as a domain-specific language and construct a segmentation-safe Question-and-Answer (QA) dataset that preserves Abstract Syntax Notation (ASN.1) structure through linearization prior to Byte Pair Encoding (BPE) tokenization. The proposed approach combines parameter-efficient adaptation with schema-bounded prompting to ensure syntactic and procedural fidelity. Evaluation introduces a standards-aware triad—ASN.1 conformance, field-level coverage analysis, and uplink-to-downlink state-machine checks—alongside semantic similarity and latency profiling across 120 configurations. On 30k 5G request–response pairs plus an additional 4.8k QA turns from 4G sessions, our 8B model achieves a median cosine similarity of 0.97, a 61% relative gain over a zero-shot baseline, while sustaining high conformance rates. These results demonstrate that LAMs, when augmented with protocol-aware reasoning, can directly orchestrate control-plane procedures, laying the foundation for the future Artificial Intelligence (AI)-native Radio Access Network (RAN).

We propose a data-driven approach for large-scale cellular network optimization, using a production cellular network in London as a case study and employing Sionna ray tracing for site-specific channel propagation modeling. We optimize base station antenna tilts and half-power beamwidths, resulting in more than double the 10%-worst user rates compared to a 3GPP baseline. In scenarios involving aerial users, we identify configurations that increase their median rates fivefold without compromising ground user performance. We further demonstrate the efficacy of model generalization through transfer learning, leveraging available data from a scenario source to predict the optimal solution for a scenario target within a similar number of iterations, without requiring a new initial dataset, and with a negligible performance loss.

In5G and beyond (5GB) networks, the notion of cell tends to blur, as a set of points-of-access (PoAs) using different technologies often cover overlapping areas. In this context, highquality decisions are needed about (i) which PoA to use when serving an end user and (ii) how to manage PoAs, e.g., how to set their power levels. To address this challenge, we present Cluster-then-Match (CtM), an efficient algorithm making joint decisions about user assignment and PoA management. Following the human-centric networking paradigm, such decisions account not only for the performance of the network, but also for the level of electromagnetic field exposure to which human bodies incur and energy consumption. Our performance evaluation shows how CtM can match the performance of state-of-the-art network management schemes, while reducing electromagnetic emissions and energy consumption by over 80%.

In next-generation networks, cells will be replaced by a collection of points-of-access (PoAs), with overlapping coverage areas and/or different technologies. Along with a promise for greater performance and flexibility, this creates further pressure on network management algorithms, which must make joint decisions on (i) PoA-to-user association and (ii) PoA management. We solve this challenging problem through an efficient and effective solution concept called Cluster-then-Match (CtM). While state-of-the-art approaches tend to focus on performance-related metrics, e.g., network throughput, CtM makes human-centric decisions, where pure network performance is balanced against energy consumption and electromagnetic field exposure. Importantly, such human-centric metrics concern all humans in the network area — including those who are not network users. Through our performance evaluation, which leverages detailed models for EMF exposure estimation and standard-specified signal propagation models, we show that CtM outperforms state-of-the-art network management schemes that solely focus on network performance, including those utilizing machine learning, reducing energy consumption by over 80% in indoor scenarios, and over 36% in outdoor ones.

We address the challenge of designing cellular networks for uncrewed aerial vehicles (UAVs) corridors through a novel data-driven approach. We assess multiple state-of-the-art high-dimensional Bayesian optimization (HD-BO) techniques to jointly optimize the cell antenna tilts and half-power beamwidth (HPBW). We find that some of these approaches achieve over 20 dB gains in median SINR along UAV corridors, with negligible degradation to ground user performance. Furthermore, we explore the HD-BO’s capabilities in terms of model generalization via transfer learning, where data from a previously observed scenario source is leveraged to predict the optimal solution for a new scenario target. We provide examples of scenarios where such transfer learning is successful and others where it fails. Moreover, we demonstrate that HD-BO enables multi-objective optimization, identifying optimal design trade-offs between data rates on the ground versus UAV coverage reliability. We observe that aiming to provide UAV coverage across the entire sky can lower the rates for ground users compared to setups specifically optimized for UAV corridors. Finally, we validate our approach through a case study in a real-world cellular network, where HD-BO identifies optimal and non-obvious antenna configurations that result in more than double the rates along 3D UAV corridors with negligible ground performance loss.

As traditional cellular base stations (BSs) are opti-mized for 2D ground service, providing 3D connectivity to un-crewed aerial vehicles (UAVs) requires re-engineering of the existing infrastructure. In this paper, we propose a new methodology for designing cellular networks that cater for both ground users and UAV corridors based on Bayesian optimization. We present a case study in which we maximize the signal-to-interference-plus-noise ratio (SINR) for both populations of users by optimizing the electrical antenna tilts and the transmit power employed at each BS. Our proposed optimized network significantly boosts the UAV performance, with a 23.4 dB gain in mean SINR compared to an all-downtilt, full-power baseline. At the same time, this optimal tradeoff nearly preserves the performance on the ground, even attaining a gain of 1.3 dB in mean SINR with respect to said baseline. Thanks to its ability to optimize black-box stochastic functions, the proposed framework is amenable to maximize any desired function of the SINR or even the capacity per area.

In this paper, we leverage a multi-agent reinforcement learning (MARL) framework to jointly learn a computation of-floading decision and multichannel access policy with corresponding signaling. Specifically, the base station and industrial Internet of Things mobile devices are reinforcement learning agents that need to cooperate to execute their computation tasks within a deadline constraint. We adopt an emergent communication protocol learning framework to solve this problem. The numerical results illustrate the effectiveness of emergent communication in improving the channel access success rate and the number of successfully computed tasks compared to contention-based, contention-free, and no-communication approaches. Moreover, the proposed task offloading policy outperforms remote and local computation baselines.

One of the consequences of network densification is more frequent handovers (HO). HO failures have a direct impact on the quality of service and are undesirable, especially in scenarios with strict latency, reliability, and robustness constraints. In traditional networks, HO-related parameters are usually tuned by the network operator, and automated techniques are still based on past experience. In this paper, we propose an approach for optimizing HO thresholds using Bayesian Optimization (BO). We formulate a multi-objective optimization problem for selecting the HO thresholds that minimize HOs too early and too late in indoor factory scenarios, and we use multi-objective BO (MOBO) for finding the optimal values. Our results show that MOBO reaches Pareto optimal solutions with few samples and ensures service continuation through safe exploration of new data points.

In this paper, we apply an multi-agent reinforcement learning (MARL) framework allowing the base station (BS) and the user equipments (UEs) to jointly learn a channel access policy and its signaling in a wireless multiple access scenario. In this framework, the BS and UEs are reinforcement learning (RL) agents that need to cooperate in order to deliver data. The comparison with a contention-free and a contention-based baselines shows that our framework achieves a superior performance in terms of goodput even in high traffic situations while maintaining a low collision rate. The scalability of the proposed method is studied, since it is a major problem in MARL and this paper provides the first results in order to address it.

We provide the reader with an accessible yet rigorous introduction to Bayesian optimisation with Gaussian processes (BOGP) for the purpose of solving a wide variety of radio resource management (RRM) problems. We believe that BOGP is a powerful tool that has been somewhat overlooked in RRM research, although it elegantly addresses pressing requirements for fast convergence, safe exploration, and interpretability. BOGP also provides a natural way to exploit prior knowledge during optimization. After explaining the nuts and bolts of BOGP, we delve into more advanced topics, such as the choice of the acquisition function and the optimization of dynamic performance functions. Finally, we put the theory into practice for the RRM problem of uplink open-loop power control (OLPC) in 5G cellular networks, for which BOGP is able to converge to almost optimal solutions in tens of iterations without significant performance drops during exploration.

In this paper, a hybrid model to compute the indoor-to-outdoor path-loss for femtocells is proposed. In this approach, indoor propagation is computed with the FDTD (Finite-Difference Time-Domain) model, thus providing high accuracy inside the building where the femtocell is deployed. On the other hand, outdoor propagation is computed with an empirical approach. Performance of the model is evaluated with on-site radio measurements, and it is shown that the proposed approach reaches a reasonable compromise: it overcomes the low prediction accuracy of using only an empirical model, while reducing the computational load of using FDTD for the whole scenario.

Femtocells are a promising solution for the provision of high indoor coverage and capacity. Furthermore, OFDMA-based femtocells have proven to be highly versatile when dealing with cross-layer co-channel interference thanks to the allocation of frequency subchannels. However, concerns still exist related to the impact of the different access methods to femtocells in an overlayed network. Femtocells based on a Closed Subscribers Group, where only device owners are allowed connectivity introduce severe interference to macrocell users. On the other hand, Open Access femtocells where any user can connect, does not bring many advantages to the femtocell owner. In this paper, an intermediate access method based on a limited access is proposed. The performance of the model is evaluated throughout system-level simulations and it is shown that limited access contributes to seriously reduce cross-layer interference while guaranteeing a minimum performance to the femtocell subscribers.

Finite-difference electromagnetic methods have been used for the deterministic prediction of radio coverage in cellular networks. This paper introduces an approach that exploits the spatial power distribution obtained from these techniques to characterize the random variations of fading due to multipath propagation. Although the presented method is equally applicable to any finite-difference algorithm able of computing electromagnetic field patterns (e.g. PSTD, ParFlow, ...), it has been exemplified here by means of FDTD. Furthermore, in order to test the reliability of this approach, the predicted fading distributions are compared against real measurements in a residential indoor-to-outdoor scenario. Finally, the practical usability of this fading prediction approach is tested throughout implementation in a WiMAX femtocells system-level simulator.

The Finite-Difference Time-Domain (FDTD) is not just one of the most accurate propagation prediction methods but also one of the most computationally intensive models, especially when simulating big areas such as urban environments. To overcome this problem, one common approach is to perform simulations at a much lower frequency than that of the real system, and to calibrate the simulation parameters to adjust the frequency response of the materials. In this study a theoretical analysis of the consequences of applying such frequency reduction is performed. To compare the theoretical study some numerical experiments are conducted and matched to the analytical results. Finally, the error of the simulations at lower frequencies is evaluated, allowing us to obtain an explanation for the observed behavior.