In next-generation (xG) wireless communication networks, developing generalizable learning models that inherently adapt to diverse conditions is crucial. This paper proposes a reinforcement learning (RL) framework for subchannel (SC) allocation in low Earth orbit satellite (LEOS)-assisted integrated access and backhauling (IAB) networks. We consider an integrated terrestrial-satellite network, where a LEOS provides backhaul services to cellular base stations (BSs) in remote areas while forwarding data from mobile user equipments (UEs) to the core network. The objective is to maximize the achievable rate for UEs while satisfying demand requirements and backhaul constraints. To ensure generalizable and efficient SC allocation across environments, we formulate the resource management problem as an invariant policy learning framework, which is decomposed into two subproblems: state representation learning and policy optimization. Our approach learns state representations that remain invariant across diverse environments. Additionally, the invariant policy, obtained from the hypergraph output layer, captures the fundamental causes of successful actions, enabling robust decision-making. By embedding problem constraints into both the model architecture and the training objective, the framework enhances the transparency of the invariant policy optimization process. Furthermore, we derive a data-dependent generalization bound that characterizes the policy’s performance in unseen environments. Simulation results demonstrate that the proposed policy consistently outperforms traditional methods across multiple environments.

The joint problem of carrier aggregation (CA), load balancing, and backhauling (JCALB) is studied in the context of non-terrestrial networks (NTNs) based on low-earth orbit satellites (LEOS). While CA can potentially enhance the communication capacity by dynamic selection of component carriers (CCs) or frequency bands for the LEOS, load balancing adjusts the portion of each CC utilized by individual LEOS in order to optimize resource utilization. Aiming to minimize the usage of each CC by individual satellites while maximizing their achievable total rate, we formulate the JCALB problem as a mixed-integer stochastic optimization problem involving both discrete and continuous decision variables, which is NP-hard. To solve the problem suboptimally, we divide it into two sub-problems: backhauling and activating/deactivating CCs for the satellites, and load balancing over the CCs. For the first subproblem, we develop a generative AI-based decision-making (GADM) algorithm based on a diffusion model. We apply the GADM algorithm to actor-critic and multi-arm bandit frameworks in reinforcement learning, in order to develop diffusion-based actor-critic CA and backhauling (DA2CAB) and diffusion-based upper confidence bound (UCB) CA and backhauling (DU2CAB) methods for LEOS-based NTNs. Finally, given the activated CCs for the satellites, we develop an iterative and distributed load balancing algorithm within NTNs. The simulation results demonstrate that our derived diffusion-based algorithms enable LEOS to achieve a higher transmission capacity while allocating fewer CCs and subchannels (SCs) compared to algorithms based on the double deep Q-Network (DDQN) and the traditional UCB approach.

Edge computing offers a promising solution to enhance network reliability. In this study, we investigate the integration of mobile edge computing (MEC) technology and unmanned aerial vehicles (UAVs) within the context of smart agriculture. Smart agriculture relies on resource-constrained Internet of Things (IoT) devices for local environmental monitoring and data collection. These IoT devices send the collected data to UAVs for analysis. A central theme of this work is the focus on the applications generated by each UAV and the consideration of their topology to derive our optimization algorithm. To tackle these challenges, we propose harnessing the computational and power resources of UAVs and MEC at the network’s edge to offload and execute resource-intensive tasks in UAV-MEC-assisted networks. Our research focuses on the joint optimization of power allocation and task offloading in these wireless networks. Central to our investigation is the problem of minimizing the energy-time cost (ETC) for the UAVs, considering the interdependencies among tasks. To address this complex problem efficiently, we introduce graph convolutional neural networks (GCNs) and reinforcement learning (RL)-based techniques. We employ a directed acyclic graph (DAG) to model task interdependencies, with GCNs characterizing the DAG. Our approach incorporates an actor-critic method with embedding layers, trained using the compound-action actor-critic (CA2C) algorithm. Our findings reveal a significant improvement in minimizing both delay and energy consumption, with a 27% percent reduction in delay and a 45% reduction in consumed energy for executing complex, interdependent tasks.

Sentiment Analysis Systems (SASs) are data-driven Artificial Intelligence (AI) systems that assign one or more numbers to convey the polarity and emotional intensity of a given piece of text. However, like other automatic machine learning systems, SASs can exhibit model uncertainty, resulting in drastic swings in output with even small changes in input. This issue becomes more problematic when inputs involve protected attributes like gender or race, as it can be perceived as bias or unfairness. To address this, we propose a novel method to assess and rate SASs. We perturb inputs in a controlled causal setting to test if the output sentiment is sensitive to protected attributes while keeping other components of the textual input, such as chosen emotion words, fixed. Based on the results, we assign labels (ratings) at both fine-grained and overall levels to indicate the robustness of the SAS to input changes. The ratings can help decision-makers improve online content by reducing hate speech, often fueled by biases related to protected attributes such as gender and race. These ratings provide a principled basis for comparing SASs and making informed choices based on their behavior. The ratings also benefit all users, especially developers who reuse off-the-shelf SASs to build larger AI systems but do not have access to their code or training data to compare.

As one of the key technologies in 5G networks, Carrier Aggregation (CA) is studied in this paper. In CA, Component Carriers (CCs) can be activated and deactivated depending on multiple factors, e.g., energy consumption and Quality of Service (QoS) demand of users. We propose CC management strategies where each User Equipment (UE) minimizes its average delay and at the same time minimizes its power consumption while considering that CCs can be activated and deactivated only at certain times, as in real-world CA implementations. We first model the problem as a centralized multi-objective optimum CC management problem. Since centralized approaches would impose a large overhead on the system, we then develop a semi-distributed solution by modeling the problem as a stochastic game and propose a multi-agent Double Deep Q-Network (DDQN) based CC management algorithm to solve the stochastic game. We finally compare the proposed approaches with single CC activation and all-CC activation baseline schemes. Simulation results show that our proposed algorithms outperform the all-CC algorithm in terms of UE power consumption and have the capability of transmitting a number of bits with delay close to the all-CC scheme. Meanwhile, our DDQN-based algorithm decreases the UE power consumption by about 20% with respect to the all-CC scheme.

We present a resource management framework for load-coupled downlink cellular OFDMA networks considering the load factor of an individual base station (BS) per resource block (RB), i.e., the number of adjacent sub-carriers (SCs), as the variable of interest in the resource management problem. The load factor of a BS per RB, which corresponds to the fraction of active SCs in the BS per RB, is an indicator of the level of resource consumption, and it affects the interference caused to that RB reused in other BSs, and thereby, results in a load-coupled OFDMA system. We first propose two distributed schemes to minimize: (i) the total load factor of the BSs (which would in turn increase the number of supportable users in the system), and (ii) the total downlink transmit power level of the BSs. Then, we derive the necessary and sufficient conditions for checking the feasibility of given target-rate requirements (also referred to as demand vector) for users. Accordingly, an iterative and distributed scheme is proposed to check the feasibility of a given demand vector. Next, for a priority-based load-coupled network, we propose a priority-based gradual removal algorithm to support the maximal number of low-priority users while satisfying the demands of the high-priority users. To evaluate the performance of our proposed schemes for resource management and admission control in load-coupled OFDMA networks, the theoretical investigations are complemented with Monte Carlo simulations.

To improve the spectral efficiency in long-term evolution systems, the resource blocks (RBs) are shared among different cells/base stations (BSs) resulting in interference among the cells/BSs on each RB, although all the sub-carriers (SCs) in an RB may not be used in a cell. Defining the load of a given BS per RB as the fraction of the active SCs in that RB, in this paper, we present a generalized signal-to-interference-and-noise-ratio (SINR) model for downlink users on a given RB. This model considers both the transmit powers of the BSs and the loads of the cells over that RB. Under this load-coupled SINR model, to study the feasibility of a given rate demand vector for users, we formulate an optimization problem of minimizing the total load of the BSs on the RBs. Then, for two different scenarios of feasible and infeasible demand vectors, respectively, we study the load management problem (i.e., minimizing the total load of the BSs on the RBs) and admission control problem (i.e., finding the sub-set of users with maximum cardinality whose demands can be concurrently satisfied), respectively. Our theoretical investigations, which provide guidelines for designing radio resource management methods for load-coupled OFDMA networks, are complemented through Monte Carlo simulations.

The problem of joint load management and admission control (JLAC) in underlay OFDMA-based cognitive radio networks (CRNs) is studied. The fraction of the active sub-carriers in a base station (BS) in each resource block (RB) is defined as the load factor of the BS per RB. In the JLAC problem, we simultaneously minimize the secondary users' outage ratio and the total load factors of the BSs via the RBs, subject to the constraint that the primary users are protected. This problem is a NP-hard problem. We first relax it into a convex optimization problem. Then, by employing the optimal solution to the relaxed JLAC problem, we derive a heuristic algorithm to gradually remove the most adversative secondary users imposing the most interference to the primary users. The performance of the proposed algorithm is studied in terms of the outage ratio of secondary users and the total load factors of the BSs via the RBs through extensive simulations.

In this paper, the region of SINRs vector for a given user on its assigned sub-carriers (SCs) which corresponds to a given target-rate, is characterized. In doing so, given the user's data rate, the boundary of region of the SINRs vector corresponding to data rate for the user on its assigned SCs is obtained. We show that for a given preference of the user, the point laid on that boundary corresponds to the minimum SINRs vectors for providing the given data rate for the user. Therefore, characterizing the region of SINRs vector corresponding to the target-rate for each users on their assigned SCs and at the same time characterizing the feasible SINRs region on each SCs (which is a traditional problem and addressed already) enable us to study if a given target-rate vector is admissible or not.