Large Language Models (LLMs) demonstrate impressive capabilities in natural language understanding and generation, but incur high communication overhead and privacy risks in cloud deployments, while facing compute and memory constraints when confined to edge devices.Cloud–edge inference has emerged as a promising paradigm for improving privacy in LLM services by retaining sensitive computations on local devices.However, existing cloud–edge inference approaches apply uniform privacy protection without considering input sensitivity, resulting in unnecessary perturbation and degraded utility even for non-sensitive tokens. To address this limitation, we propose Privacy-aware Routing for Inference with Semantic Modulation (PRISM), a context-aware framework that dynamically balances privacy and inference quality. PRISM executes in four stages: (1) the edge device profiles entity-level sensitivity; (2) a soft gating module, also on the edge, selects an execution mode -cloud, edge, or collaboration; (3) for collaborative paths, the edge applies adaptive two-layer local differential privacy based on entity risks; and (4) the cloud LLM generates a semantic sketch from the perturbed prompt, which is then refined by the edge-side small language model (SLM) using local context.Our results show that PRISM consistently achieves superior privacy-utility trade-offs in various scenarios, reducing energy consumption and latency to 40–50% of baseline methods such as Uniform and Selective LDP, while maintaining high output quality under strong privacy constraints. These findings are validated through comprehensive evaluations involving realistic prompts, actual energy measurements, and heterogeneous cloud–edge model deployments.

In this paper, we address the novel problem of jointly considering dynamic function configuration, function download latency, multi-slot task offloading, and energy harvesting in order to process the maximum number of tasks. To this end, we formulate a mixed integer linear program (MILP) that optimizes (i) task offloading of servers, (ii) function configuration, (iii) function scheduling, (iv) function download decision, (v) energy usage, (vi) channel allocation. It provides a theoretical optimum, but leverages non-causal information and becomes computationally intractable in large scale networks. To address this challenge, we propose a fully distributed protocol called Distributed Task Offloading with Energy Harvesting (D-TOEH), where each server establishes a function importance ranking to activate functions required by more tasks. Our simulation results show that, on average, the performance of D-TOEH attains 71.36% of the theoretical optimum while ensuring low polynomial time complexity.

Future Internet of things (IoT) networks will host applications that involve data collection and computation tasks on one or more servers. To this end, this paper proposes the first mixed integer linear program (MILP) to schedule and embed applications on energy harvesting nodes, where it optimizes (i) the sampling time of devices, (ii) whether to run an application, and (iii) the energy usage of devices, gateways and servers. To ensure applications are run often, we adopt the maximum age of service (AoS) metric, and set the MILP’s objective to minimize the maximum AoS or min-max AoS of applications. This paper also proposes two novel solutions: (i) a receding horizon control (RHC) based method, and (ii) a solution that greedily embeds applications according to their AoS. The results show that the min-max AoS of RHC and greedy approach is respectively 1.07x and 1.13x higher than MILP.

A key issue when operating an Internet of things (IoT) network is that devices may be infected by malware. Consequently, when collecting data from these devices, e.g., using an uncrewed aerial vehicle (UAV), malware data may be transmitted to a gateway that is then used to attack servers. To address this issue, we equip a UAV with a virtual network function, and study a novel problem involving the allocation of computation and communication resource to identify malware traffic from solar-powered devices. To solve this problem, we first outline a mixed integer linear program (MILP) that jointly optimizes i) UAV placement, ii) UAV trajectory, iii) data processing, iv) channel allocation, and v) energy usage of devices. To facilitate online decision making, we design a neural network-based solution called neural network mapping (NNM) to store integer valued decision variables offline. During flight, the UAV then uses the neural network to retrieve the corresponding integer values for a given scenario, which allows the UAV to solve the said MILP as a linear program that can be computed quickly. Our simulation results show that NNM is able to effectively reduce the amount of data from malware that arrives at a gateway.

This paper presents a robust Model Predictive Control (MPC) framework for trajectory tracking of unmanned aerial vehicles (UAVs), enhanced by a reinforcement learning (RL) policy for warm-start initialization and a sliding mode observer (SMO) for disturbance estimation. The proposed architecture addresses multifaceted performance degradation, including residual trajectory tracking errors, reduced control responsiveness, and slow early-stage convergence, primarily caused by external disturbances, model uncertainties, and time-varying environmental conditions. An enhanced adaptive super-twisting SMO is developed to estimate lumped disturbances in real time. These estimates are integrated into the MPC prediction model to compensate for mismatches between nominal dynamics and true system behavior. To accelerate control convergence, particularly in improving early stage performance, an offline RL warm-start policy is used to generate an initial control sequence for MPC. The overall framework preserves the constraint handling and predictive capabilities of conventional MPC, while significantly improving robustness, stability, and tracking accuracy. The simulation results validate the effectiveness of the proposed method in achieving reliable and precise trajectory tracking under challenging and uncertain operating scenarios.

In this article, we consider an unmanned aerial vehicle (UAV) with an in-band full-duplex radio that is used to interconnect industrial Internet of things (IIoT) devices and exploit their computation and energy resources to help process data. We formulate a mixed integer linear program to optimize the first sampling rate of each device, second amount of data the UAV transmits and receives to/from a device, third position of the UAV over time, and finally the number of virtual machines used by devices and UAV to compute data. We also propose a distributed protocol to determine quantity using aforementioned points. Our results show that an IBFD-UAV has a higher max–min computed sampling rate as compared to when the UAV uses a half duplex radio. Moreover, the said protocol achieves a max–min rate that is 80% optimal.

Serverless computing has been widely regarded as an ideal computing paradigm, enabling edge servers to host serverless functions. Due to its high scalability and usage-based pricing model, it provides efficient services across various applications. However, in the deployment process of serverless applications, past works lack considerations for the parallel relationships between stateful functions, which increases end to end latency. To leverage the parallel dependencies between functions, we propose a strategy for dependent function parallelization deployment, named LPCD (Low latency Parallel Candidate Deployment strategy). By partitioning the problem into inter-layer function deployment and analyzing optimal substructures, a heuristic algorithm is introduced to determine candidate deployment strategies for each layer of the users, which aims at identifying the optimal edge server for each function instance during deployment to enhance user satisfaction. Through simulation experiments, we evaluate the performance of the strategy. The experiments results indicate that the average latency was reduced by at least 41% compared with the state-of-the-art strategies.

In this study, we design novel approximately mutually unbiased bases (AMUBs) sequences with arbitrary lengths and large family sizes for massive connection systems. Sequences with low correlations are highly demanded for many wireless communications systems, including Industrial Internet of Things (IIoT) systems for various applications. While many sets of sequences have been designed in the past decades, the requirement of large family size, i.e., the number of available sequences for massive connection systems has not yet been addressed. It is well known that mutually unbiased-based (MUB) sequences process desired correlation properties with large family sizes. However, the family size based on the current construction methods is limited by the length of the MUB sequences. In real applications, the longer length may lead to higher overhead and affect the overall transmission rate. This drawback makes MUB sequences have limited applications for industrial massive connection systems. In this article, we modified the original sequences generator of MUB from a quadratic polynomial to a cubic polynomial to further increase the family size. To generate AMUB sequences with arbitrary lengths, we then proposed a construction method based on the exponential sums over finite fields, and optimized the continuous peak-to-average power ratio (PAPR) of the proposed AMUB sequences for real applications. Given the dimension of MUB M, the proposed method can increase M times the number of available sequences. Meanwhile, the length restrictions in MUB sequence construction are removed. Theoretical cross-correlation (CC) bounds are provided and show low correlations of the proposed sequences. The low CC, PAPR, and increased family size of the proposed sequences are then verified by numerical results.

This article considers routing and link scheduling in a two-tier wireless backhaul network. The first tier consists of routers and the second tier consists of radio frequency (RF)-energy-harvesting Internet-of-Things (IoT) devices that rely on routers for energy. Our aim is to derive the shortest time division multiple access (TDMA) link schedule that satisfies the traffic demand of routers and energy demand of IoT devices. We formulate a linear program (LP) to jointly derive a routing and link schedule solution. We also propose a heuristic link scheduler called transmission set generation (TSG) to generate transmission sets and to derive the transmit power allocation of routers. In addition, we present a novel routing metric that considers RF-energy-harvesting devices on a given path. TSG on average achieves 31.25% shorter schedules as compared to competing schemes. Finally, our novel routing metric results in link schedules that are at most 24.75% longer than those computed by LP.

Virtualization of devices operating in Internet of Things (IoT) networks allows them to host functions or tasks from different users; these devices can thus execute multiple on demand sensing and data processing services concurrently. Devices, however, have limited energy and operational lifetime. To this end, this article considers supporting virtual network functions (VNFs) in a radio-frequency (RF)-charging network with a hybrid access point (HAP). Our aim is to minimize the energy used by the HAP to power devices in order to support deployed VNFs. It outlines a mixed-integer linear program (MILP) to jointly optimize VNFs placement, routing and link scheduling, and also the HAP’s charging duration. Furthermore, it proposes a heuristic, called decoupled greedy algorithm (DGA), that first assigns VNFs onto devices with the highest energy level before optimizing the HAP’s charging, routing, and link schedule. Our results show that DGA has a probability higher than 0.95 to successfully serve a service request. Furthermore, DGA consumes up to 36.43% less energy than competing methods.

Many applications operating in the Internet of Things (IoT) require timely and fair data collection from devices. This has motivated research into a new metric called Age of Information (AoI). This paper contributes to this effort by proposing to minimize the maximum average AoI (min-max AoI) in a multi-hop IoT network comprising of solar-powered Power Beacons (PBs). It outlines a Mixed Integer Linear Program (MILP) that jointly optimizes: (i) the beamforming vector used by PBs to charge devices, and (ii) routing, which determines how samples from devices are forwarded to a sink node, and (iii) the sampling time of sources. It also presents two protocols: Centralized Linear Relaxation (CLR) and Distributed Path Selection (DPS), respectively. CLR is run by the sink to determine the transmit power of PBs and the path of each source using two Linear Programs (LPs). On the other hand, DPS is a distributed approach whereby PBs and sources make their own decisions using local information. Our simulation results show that min-max AoI increases with the number of sources, but reduces with increasing number of PBs. The number of paths available to a source, the number of frames, and solar panel size have limited impact on performance. The min-max AoI of CLR and DPS is$1.60\times $and$1.95\times $higher than that of MILP.

This article considers a novel Internet of Things network comprising of sensor devices and power beacons (PBs); both types of nodes are equipped with a cognitive radio (CR). In addition, these sensor devices are powered by radio-frequency signals from PBs. Our aim is to maximize the minimum rate of devices acting as sources. We outline the first mixed integer linear program (MILP) that jointly optimizes the channel assignment of PBs and devices, beamforming vector of PBs, data routing over multiple hops and link activation schedule of devices. We also design a distributed protocol called distributed max–min rate with CR (D-MRCR) for use by devices and PBs. Devices set their operation mode using local information and use a game theory-based approach to iteratively adjust their transmit power. On the other hand, each PB employs a linear program to determine its beamforming vector. Our results show that the max–min rate of D-MRCR is within 51.84% that of MILP.

Future Internet-of-Things (IoT) networks will consist of radio-frequency (RF) energy harvesting devices that are charged by solar-powered power beacons (PBs). To this end, this article aims to maximize the minimum data rate of devices acting as sources operating in a multihop IoT network. The main problem is to decide the amount of energy delivered by solar-powered PBs, routing of data from each source, and link scheduling, which determines the capacity of links. To this end, we make two contributions. First, we present a linear program (LP) to optimize the max-min rate of sources. Our LP considers nonlinear RF conversion at devices, energy storage loss at devices due to the imperfect battery, and time-varying channel quality, which affect the amount of energy harvested by devices. The second contribution is a novel distributed protocol called distributed max-min rate allocation (D-MRA), whereby devices only need local information, such as their battery and data buffer state to make decisions. Our results show that the max-min rate of D-MRA is 58.25% that of LP, which requires global information, in all tested cases.

We consider Rechargeable Wireless Sensor Networks (R-WSNs) where nodes harvest energy from both solar and the Radio Frequency (RF) transmissions of their neighbors. Our aim is to maximize the minimum source or sensing rate of nodes. This rate is determined by the available energy at sensor nodes as well as link capacity, which is determined by the set of transmitting nodes. In this paper, we first study and show the benefits of energy sharing. Intuitively, a sensor node should share its energy if doing so increases source rates. We present a novel Linear Program (LP) to determine the routing, link schedule, energy transmission, and reception time that maximize the minimum source rate of a given R-WSN. Our numerical results indicate that, on average, the minimum transmission rate of sensor nodes increased by 16.03 percent when nodes share energy. This motivates the development of a practical protocol called E-RSVP that iteratively increases the time slots of each source node. It also considers using time slots for transmission or reception of energy. Our simulation results show E-RSVP yields minimum source rates that are 14.80 percent higher as compared to the case without energy sharing.

In a rechargeable wireless sensor network (rWSN), the amount of data forwarded by source nodes to one or more sinks is bounded by the energy harvesting rate of sensor nodes. To improve sensing quality, we consider a novel approach whereby we place a finite number of auxiliary chargers (ACs) with wireless power transfer and energy harvesting ability to boost the energy harvesting rate of some sensor nodes. We formulate a mixed integer linear program (MILP) to determine the subset of nodes that if upgraded will maximize the minimum source rate. We also propose two heuristic algorithms to place ACs in large-scale rWSNs: greedy node deployment (GND), which checks every nonupgraded sensor node and places an AC next to the one yielding the highest increase in max-min rate; and one-unit energy deployment (OUED), which uses a relaxed version of the MILP to first share one unit of energy among sensor nodes. It then upgrades the sensor node with the highest one-unit share. Our results show that the max-min rate obtained by GND and OUED is, respectively, within 99.60% and 97.82% of the max-min rate derived by MILP in small networks with at most 90 nodes. In large networks with 200 nodes, the maximum gap between OUED and GND is only 0.191 kb/s. Lastly, OUED runs at least five times faster than GND.