Many decentralized distillation methods are designed around training-time coordination, yet deploy each node in isolation even when more capable neighbors remain available at inference time. This is an incomplete objective for settings such as IoT, where devices are heterogeneous, data is scarce and skewed, and a node's strongest neighbors may far exceed its own local capacity. We study how nodes should train so that their predictions compose well at deployment, and how each node should learn whom to trust. Under a server-free, model-agnostic protocol where nodes exchange only queries and soft predictions, we propose Learned Neighbor Trust (LNTrust) wherein each node learns a compact trust function over its neighborhood from local validation evidence. This trust function gates auxiliary distillation during training and defines a deployment ensemble at inference, so that collaboration learned during training transfers directly to deployment. Across datasets and topologies, LNTrust improves deployed accuracy over the strongest output-only baseline by large margins while using significantly less communication than previous methods.

We examine a simplified distributed collaborative beamforming (DCB) system consisting of two transmitters sending data to a single receiver over independent channels with significant time-varying phase distortion. The transmitters sample the state of their respective channels using a periodic sounding waveform sent by the receiver and use the last sensed state along with the sample age to correct for the channel phase distortion. As the age of the last sampled channel state increases, the transmitters are correctly estimating the current state with decreasing likelihood, leading to reduced beamforming gain. However, because sensing and transmitting must occur on the same channel, the transmitters cannot both send data and sample the channel at the same time. As a result, the cost associated with channel sensing is the transmitters not sending data for some duration; this compromise is embodied by the frame-averaged expected beamforming gain. A general model is developed using a quantized representation of the channel phase space. Simulations are used to characterize the error incurred by using a discrete phase representation over a continuous phase, and optimal sensing period is calculated across a range of model parameters.

The growing need for privacy-preserving and communication-efficient machine learning in dynamic wireless environments has catalyzed the development of federated learning (FL). While traditional FL schemes suffer from straggler effects or significant model diversity, this paper proposes a novel semi-asynchronous strategy named Wait-for-Optimal-Set (WOS). WOS dynamically determines both the cut-off time and the optimal subset of clients to participate in each FL round based on their computation latency and model-version gap. A dynamic resource block (RB) allocation algorithm is further integrated to enable immediate transmission upon training completion, thereby reducing transmission delay. Through extensive simulations in a UAV-autonomous team-assisted FL scenario, WOS demonstrates superior convergence speed and model accuracy compared to existing state-of-the-art methods, especially under time-varying channels and heterogeneous device capabilities.

In a coherent communication system consisting of an open-loop distributed transmit array sending messages to a distributed receive array, the combined transmit-receive gain is characterized by the coherent communication gain. We consider the problem of optimizing coherent communication gain using the positions of the individual transmitter and receiver nodes as well as the beam angle of the transmit array as degrees of freedom. We focus on the use of gradient descent to find locally optimal configurations for node positions, which is motivated by two observations: first, the NP-hardness of the problem precludes an exhaustive search for the globally optimal configuration of node positions; and second, the positions of the network nodes are likely not arbitrary. That is, the initial, non-optimized node placement is intentional and is determined by higher-layer network objectives. The hypothesis is that the coherent communication gain of a communication network can be improved in a deterministic fashion using a steepest descent algorithm to make relatively small adjustments to node positions. We develop the closed-form expressions for the rate of change of coherent communication gain with respect to node positions and transmit array beam angle. Next, we use the expressions to implement a spherical quadratic steepest descent (SQSD) algorithm and use simulations to test SQSD alongside pattern search and particle swarm optimization to determine theoretical gain improvements achieved by the algorithms, as well as the expected average node displacement.

Age of Information (AoI) is a metric that can be used to measure the freshness of information. Since its inception, there have been active research efforts on designing scheduling algorithms to AoI-related problems. These problems vary in specific AoI-based objectives and network settings. For each problem, typically a custom-designed scheduler was developed. Instead of following the (custom-design) path, we envision and pursue a general framework that can be applied to design a wide range of schedulers to solve AoI-related problems. As a first step toward this vision, we present a general framework-Eywa, that can be applied to construct high-performance schedulers for a family of AoI-related optimization and decision problems, all sharing a common setting of an IoT data collection network. We show how to apply Eywa to solve three important problems: 1) to minimize weighted sum of AoI; 2) to minimize bandwidth requirement under AoI constraints; and 3) to determine the existence of feasible schedulers to satisfy AoI constraints. We show that for each problem, Eywa can either offer a stronger performance guarantee than the state-of-the-art algorithms or provide new (or general) results that are not available in the literature.

5G/6G sidelink communications addresses the challenge of connecting outer UEs, which are unable to directly access a base station (gNodeB), through inner UEs that act as relays to connect to the gNodeB. The key performance indicators include the achievable rates, the number of outer UEs that can connect to a gNodeB, and the latency experienced by outer UEs in establishing connections. We consider problem of determining the assignment of outer UEs to inner UEs based on the channel, interference, and traffic characteristics. We formulate an optimization problem to maximize a weighted sum rate of UEs, where weights can represent priority, waiting time, and queue length. This optimization accommodates constraints related to channel and interference characteristics that influence the rates at which links can successfully carry assigned traffic. While an exhaustive search can establish an upper bound on achievable rates by this non-convex optimization problem, it becomes impractical for larger number of outer UEs due to scalability issues related to high computational complexity. To address this, we present a greedy algorithm that incrementally selects links to maximize the sum rate, considering already activated links. This algorithm, although effective in achieving high sum rates, may inadvertently overlook some UEs, raising concerns about fairness. To mitigate this, we introduce a fairness-oriented algorithm that adjusts weights based on waiting time or queue length, ensuring that UEs with initially favorable conditions do not unduly disadvantage others over time. We show that this strategy not only improves the average admission ratio of UEs but also ensures a more equitable distribution of service among them, thereby providing a balanced and fair solution to sidelink communications.

Reactive jammers pose a severe security threat to robotic-swarm networks by selectively disrupting inter-agent communications and undermining formation integrity and mission success. Conventional countermeasures such as fixed power control or static channel hopping are largely ineffective against such adaptive adversaries. This paper presents a multi-agent reinforcement learning (MARL) framework based on the QMIX algorithm to improve the resilience of swarm communications under reactive jamming. We consider a network of multiple transmitter-receiver pairs sharing channels while a reactive jammer with Markovian threshold dynamics senses aggregate power and reacts accordingly. Each agent jointly selects transmit frequency (channel) and power, and QMIX learns a centralized but factorizable action-value function that enables coordinated yet decentralized execution. We benchmark QMIX against a genie-aided optimal policy in a no-channel-reuse setting, and against local Upper Confidence Bound (UCB) and a stateless reactive policy in a more general fading regime with channel reuse enabled. Simulation results show that QMIX rapidly converges to cooperative policies that nearly match the genie-aided bound, while achieving higher throughput and lower jamming incidence than the baselines, thereby demonstrating MARL's effectiveness for securing autonomous swarms in contested environments.

We examine a simplified coherent distributed collaborative beamforming (DCB) link consisting of two transmitters sending a single data stream to a receiver. The novel elements introduced are a method for computing expected link throughput and an analysis of performance when operating over wireless channels with significant time-varying phase distortion. To offset the channel phase distortion, each transmitter samples the state of its channel using a periodic sounding waveform sent by the receiver. As the age of the last sampled channel state increases, so does the probability of phase error between the beamforming components. A quantized phase space model is introduced for the wireless channel and the DCB gain performance is analyzed with respect to age of channel state information (AoCSI). Additionally, we introduce a framework for computing expected throughput of the DCB link, whereby the channel sensing period is optimized to the wireless channel conditions. Future network planning for WSNs or MANETs can use this framework for incorporating DCB links. Adding DCB links to a wireless network does not necessarily require additional hardware and provides benefits such as greater link ranges, increased throughput, reduced transmit power, and reduced interference.

Recent advances in AI and machine learning have been driven by the emergence of foundation models, large pre-trained architectures capable of solving a wide range of downstream tasks with minimal task-specific tuning. These models, which learn unified representations from diverse, large-scale datasets, have transformed domains such as language, vision, and biology. However, RF signal processing systems have yet to fully benefit from this paradigm shift. Traditional approaches remain focused on closed-set, single-task objectives such as automatic modulation classification or emitter identification, limiting adaptability in dynamic scenarios that demand flexible, general-purpose reasoning across unseen waveforms and channel conditions. In this work, we introduce SemanticRF, a foundation model for RF scene analysis trained via multimodal contrastive learning to produce a unified signal-text embedding. Pretrained on a large-scale synthetic dataset spanning diverse modulation types, SNRs, and channel models, SemanticRF delivers strong performance on traditional RFML benchmarks but also demonstrates strong generalization to novel tasks such as RF scene tagging, highlighting a step towards adaptable foundation models in the RF domain.

We present a retrieval-augmented question answering framework for 5G/6G networks, where the Open Radio Access Network (O-RAN) has become central to disaggregated, virtualized, and AI-driven wireless systems. While O-RAN enables multi-vendor interoperability and cloud-native deployments, its fast-changing specifications and interfaces pose major challenges for researchers and practitioners. Manual navigation of these complex documents is labor-intensive and error-prone, slowing system design, integration, and deployment. To address this challenge, we adopt Contextual Retrieval-Augmented Generation (Contextual RAG), a strategy in which candidate answer choices guide document retrieval and chunk-specific context to improve large language model (LLM) performance. This improvement over traditional RAG achieves more targeted and context-aware retrieval, which improves the relevance of documents passed to the LLM, particularly when the query alone lacks sufficient context for accurate grounding. Our framework is designed for dynamic domains where data evolves rapidly and models must be continuously updated or redeployed, all without requiring LLM fine-tuning. We evaluate this framework using the ORANBenchmark-13K dataset, and compare three LLMs, namely, Llama3.2, Qwen2.5-7B, and Qwen3.0-4B, across both Direct Question Answering (Direct Q A) and Chain-of-Thought (CoT) prompting strategies. We show that Contextual RAG consistently improves accuracy over standard RAG and base prompting, while maintaining competitive runtime and CO2 emissions. These results highlight the potential of Contextual RAG to serve as a scalable and effective solution for domain-specific Q A in ORAN and broader 5G/6G environments, enabling more accurate interpretation of evolving standards while preserving efficiency and sustainability.

We study an Age-of-Information (AoI) scheduling problem where users can tolerate occasional violations of AoI for each source at the base station. Each user's AoI is associated with a violation tolerance constraint. We are interested in determining whether a set of users, each with a given AoI deadline, a violation tolerance constraint, and a packet loss rate (due to channel condition) is schedulable, and if so, find a feasible scheduler. For this problem, we study two cases: 1) the stable tolerant case where the tolerance rate is higher than the packet loss rate for each source and 2) the unstable tolerant case where the tolerance rate is lower than the packet loss rate for at least one source. For the stable tolerant case, we design an algorithm called stable tolerant scheduler (STS), which can find a feasible scheduler for any network when the system load is no greater than ln 2 (roughly 70%). When the system load is between ln 2 and 1, we offer a necessary and sufficient condition for STS to find a feasible scheduler by solving an optimization problem. Likewise, for the unstable tolerance case, we develop a scheduler called unstable tolerant scheduler (UTS) and its corresponding schedulability conditions. Through extensive simulations, we show that STS and UTS match our theoretical results.

In connected and autonomous vehicles, machine learning for safety message classification has become critical for detecting malicious or anomalous behavior. However, conventional approaches that rely on centralized data collection or purely local training face limitations due to the large scale, high mobility, and heterogeneous data distributions inherent in inter-vehicle networks. To overcome these challenges, this paper explores Distributed Federated Learning (DFL), whereby vehicles collaboratively train deep learning models by exchanging model updates among one-hop neighbors and propagating models over multiple hops. Using the Vehicular Reference Misbehavior (VeReMi) Extension Dataset, we show that DFL can significantly improve classification accuracy across all vehicles compared to learning strictly with local data. Notably, vehicles with low individual accuracy see substantial accuracy gains through DFL, illustrating the benefit of knowledge sharing across the network. We further show that local training data size and time-varying network connectivity correlate strongly with the model's overall accuracy. We investigate DFL's resilience and vulnerabilities under attacks in multiple domains, namely wireless jamming and training data poisoning attacks. Our results reveal important insights into the vulnerabilities of DFL when confronted with multi-domain attacks, underlining the need for more robust strategies to secure DFL in vehicular networks.

This paper studies the problem of mitigating reactive jamming, where a jammer adopts a dynamic policy of selecting channels and sensing thresholds to detect and jam ongoing transmissions. The transmitter-receiver pair learns to avoid jamming and optimize throughput over time (without prior knowledge of channel conditions or jamming strategies) by using reinforcement learning (RL) to adapt transmit power, modulation, and channel selection. Q-learning is employed for discrete jamming-event states, while Deep Q-Networks (DQN) are employed for continuous states based on received power. Through different reward functions and action sets, the results show that RL can adapt rapidly to spectrum dynamics and sustain high rates as channels and jamming policies change over time.

In emerging networked systems, mobile edge devices such as ground vehicles and unmanned aerial system (UAS) swarms collectively aggregate vast amounts of data to make machine learning decisions such as threat detection in remote, dynamic, and infrastructure-constrained environments where power and bandwidth are scarce. Federated learning (FL) addresses these constraints and privacy concerns by enabling nodes to share local model weights for deep neural networks instead of raw data, facilitating more reliable decision-making than individual learning. However, conventional FL relies on a central server to coordinate model updates in each learning round, which imposes significant computational burdens on the central node and may not be feasible due to the connectivity constraints. By eliminating dependence on a central server, distributed federated learning (DFL) offers scalability, resilience to node failures, learning robustness, and more effective defense strategies. Despite these advantages, DFL remains vulnerable to increasingly advanced and stealthy cyberattacks. In this paper, we design sophisticated targeted training data poisoning and backdoor (Trojan) attacks, and characterize the emerging vulnerabilities in a vehicular network. We analyze how DFL provides resilience against such attacks compared to individual learning and present effective defense mechanisms to further strengthen DFL against the emerging cyber threats.

This paper presents a complete signal-processing chain for multistatic integrated sensing and communications (ISAC) using 5G Positioning Reference Signal (PRS). We consider a distributed architecture in which one gNB transmits a periodic OFDM-PRS waveform while multiple spatially separated receivers exploit the same signal for target detection, parameter estimation and tracking. A coherent cross-ambiguity function (CAF) is evaluated to form a range-Doppler map from which the bistatic delay and radial velocity are extracted for every target. For a single target, the resulting bistatic delays are fused through nonlinear least-squares trilateration, yielding a geometric position estimate, and a regularized linear inversion of the radial-speed equations yields a two-dimensional velocity vector, where speed and heading are obtained. The approach is applied to 2D and 3D settings, extended to account for time synchronization bias, and generalized to multiple targets by resolving target association. The sequence of position-velocity estimates is then fed to standard and extended Kalman filters to obtain smoothed tracks. Our results show high-fidelity moving-target detection, positioning, and tracking using 5G PRS signals for multistatic ISAC.

A fundamental challenge for Citizen Broadband Radio Service (CBRS) is how to achieve optimal channel allocation for the Priority Access License (PAL) and General Authorized Access (GAA) users while protecting the incumbent federal users. To address this challenge, we develop a mathematical framework, which encompasses PAL channel allocation, PAL interference protection, GAA channel allocation, and interference protection for Environmental Sensing Capability (ESC) and incumbents per FCC guidelines. Special treatment is given to the so-called spectrum contiguity and geographical contiguity requirements for PAL channel allocation. Through linearization of the nonlinear constraints, we reformulate the original problem into a mixed-integer linear program (MILP) with no relaxation error. We conduct extensive simulation experiments on real-world data for counties along Virginia’s east coast and show that our optimal solution can offer guaranteed interference protection to the federal incumbents and PAL holders while maximizing channel utilization for the GAA users. Our simulation study also indicates that although the FCC ruling on channel contiguity and geographical contiguity requirements on PAL holders leads to some performance loss in channel allocation, such a loss is in fact limited.

This paper introduces a new theoretical framework for optimizing second-order behaviors of wireless networks. Unlike existing techniques for network utility maximization, which only consider first-order statistics, this framework models every random process by its mean and temporal variance. The inclusion of temporal variance makes this framework well-suited for modeling Markovian fading wireless channels and emerging network performance metrics such as age-of-information (AoI) and timely-throughput. Using this framework, we sharply characterize the second-order capacity region of wireless access networks. We also propose a simple scheduling policy and prove that it can achieve every interior point in the second-order capacity region. To demonstrate the utility of this framework, we apply it to an unsolved network optimization problem where some clients wish to minimize AoI while others wish to maximize timely-throughput. We show that this framework accurately characterizes AoI and timely-throughput. Moreover, it leads to a tractable scheduling policy that outperforms other existing work.

In this work, we study the tradeoff between total throughput of a network and the cumulative emitted energy experienced by one or more external nodes that are not members of the network. We formulate the linear program to determine the routing and scheduling that maximizes the fair throughput for uplink and downlink traffic subject to the energy at each external node being below a threshold. Due to the spatial reuse with simultaneous transmitters and the multi-path routing approach, the number of variables in the linear program is exponential in the number of nodes in the network, so we solve this large linear program using an iterative approach known as column generation. We present numerical results in the form of energy heatmaps generated by the optimal schedule for individual network realizations, as well as the average energy heatmap over many Monte Carlo realizations, demonstrating the impact of the external node locations. We also present throughput results as a function of the node distance and the energy threshold for the single external node case. These results characterize the tradeoff between throughput and the distance and sensed energy sensitivity level, which provides a bound with which to compare the performance of practical routing and scheduling algorithms.

The development of a family of data-driven methods, called dynamic mode decomposition, for modeling the behavior of dynamical systems through the approximation of the associated Koopman operator, has led to a rapid increase in the related research. Separately, the modeling and algorithm development for target detection in the presence of sea clutter often involves probability density function descriptions of the amplitude process, which ignores time dependency in data. In this article, we combine these data-driven methods with a stochastic differential equation model of radar scattering from the sea surface for the purpose of sea-state change detection. This approach relies on building a dynamic model of sea clutter directly from the radar measurements, without the need to estimate parameters of underlying equations. Using this model, an anomaly detection scheme is demonstrated using a Kalman filtering approach constructed from the Koopman model that is able to identify changes in the sea state.

We examine a simplified distributed collaborative beamforming (DCB) system with two transmitters sending data over independent channels with significant time-varying phase distortion. The transmitters sample the state of their respective channels using a periodic sounding waveform sent by the receiver and use the last sensed state along with the sample age to correct for the channel phase distortion. As the age of the last sampled channel state increases, the transmitters are correctly estimating the current state with decreasing likelihood, leading to reduced beamforming gain. However, because sensing and transmitting must occur on the same channel, the transmitters cannot both send data and sample the channel at the same time. As a result, the cost associated with channel sensing is the transmitters not sending data for some duration; this compromise is embodied by the frame-averaged expected beamforming gain. Expressions for instantaneous and averaged expected gain are developed, and an example is presented to demonstrate finding the optimal sensing period and how the optimal sensing period can change depending on the last sensed states of the channels.