Knowledge Graphs (KGs) store structured information but typically require specialized query languages, such as Cypher for Neo4j, creating accessibility challenges for users unfamiliar with graph syntax. Large Language Models (LLMs) offer a solution by translating natural language into Cypher queries. However, existing models—including large-scale LLMs (e.g., ChatGPT) and smaller open-source models (e.g., Llama-7B, 8B) often struggle with accurately generating domain-specific queries due to inadequate alignment with KG schemas and limited domain-specific training data. To address these limitations, we propose a training pipeline tailored specifically for domain-aligned Cypher query generation, emphasizing usability for smaller-scale models. Our method integrates template-based synthetic data generation for diverse, high-quality training samples. We combine supervised fine-tuning with preference learning to enhance domain knowledge and Cypher syntax understanding. Additionally, our approach includes a context-aware retrieval mechanism that dynamically incorporates relevant schema elements at inference, improving alignment with domain-specific knowledge. We evaluated our method on the Hetionet biomedical KG using a benchmark dataset of 240 queries across three complexity levels. Our results show that our context-aware prompting achieves a substantial improvement, increasing component matching accuracy by 23.6% for ChatGPT-4o over the vanilla prompt baseline. When applying our full training pipeline to smaller-scale models, CodeLlama-13B* achieves an execution accuracy of 69.2%, nearly matching ChatGPT-4o’s 72.1%. Importantly, our approach significantly narrows the performance gap, enabling smaller models to effectively manage complex, domain-specific tasks previously dominated by larger models. These findings demonstrate that our method is scalable, computationally efficient, and robust for practical Cypher query generation applications.

To accommodate ever-growing traffic, network operators are actively deploying high-degree reconfigurable optical add/drop multiplexers (ROADMs) to build large-capacity optical networks. High-degree ROADM-based optical networks have multiple parallel fibers between ROADM nodes, requiring the adoption of ROADM nodes with a large number of inter-/intra-node components. However, this large number of inter-/intra-node optical components in high-degree ROADM networks increases the likelihood of multiple failures simultaneously, and calls for novel methods for accurate localization of multiple failed components. To the best of our knowledge, this is the first study investigating the problem of multi-failure localization for high-degree ROADM-based optical networks. To solve this problem, we first provide a description of the failures affecting both inter-/intra-node components, and we consider different deployments of optical power monitors (OPMs) to obtain information (i.e., optical power) to be used for automated multi-failure localization. Then, as our main and original contribution, we propose a novel method based on a rules-informed neural network (RINN) for multi-failure localization, which incorporates the benefits of both rules-based reasoning and artificial neural networks (ANN). Through extensive simulations and experimental demonstrations, we show that our proposed RINN algorithm can achieve up to around 20% higher localization accuracy compared to baseline algorithms, incurring only around 4.14 ms of average inference time.

Edge computing has promoted various applications driven by artificial intelligence (AI). However, upgrading AI models during system operation may change resource and performance features. Then, the service management controller (SMC) faces an unprecedented environmental condition and has limited prior knowledge, resulting in high probabilities of policy mismatches. With the proliferation of AI applications, it is an urgent necessity that SMCs can adapt to different conditions to ensure quality of service (QoS) and resource efficiency. Therefore, this paper studies the problem of dynamic edge AI service adaptation and formulates it as a multi-task scenario adaptation problem. After that, we proposed an approach based on off-policy meta-reinforcement learning and digital twin (DT) technology. The DT system emulates a set of encountered conditions, and a meta-policy is obtained by interacting with these DTs. The executed policy is initialized as the meta-policy once AI models are upgraded. Then, it adapts to new service conditions by drawing salient information from limited transition contexts collected from a newly encountered environmental condition. Simulation results reveal that our approach can optimize QoS and adapt to different service situations.

Addressing the pressing requirement for dynamic and intelligent allocation of slicing resources, the dynamic provisioning of resources based on traffic predictions has emerged. Although this method favours proactive scheduling of network slices, more complexities are introduced by the prediction uncertainty. In addition, because multi-tenant networks are always changing in terms of technology and business model, profit-aware network slicing is becoming an important topic of study in the field of resource provision. This paper focuses on profit-aware slicing resource provisioning amid traffic uncertainty in multi-tenancy flexible Ethernet over wavelength division multiplexing networks. Specifically, we develop a profit model for multi-tenant network slicing, accounting for the impact of network prediction uncertainty, and formulate the problem as maximizing the profit of users primarily. To solve this problem, we propose a profit-aware resource provisioning approach that first checks if the slice requests are made by pruning algorithms and then determines the service relationship between slices and tenants by matching games. Simulation results demonstrate the superiority of the proposed algorithm over benchmarks in terms of user profit, total benefit, and denial ratio of service.

Satisfying users' requests based on the service level agreements of network slices is one of the most basic and vital topics of network slicing in 6G networks, and anomaly detection is regarded as a key technique for locating the abnormal status of slices. However, current studies on slice anomaly detection mostly focused on real-time monitoring of slices and ignored the prediction of potential anomalies. Generally, when anomalies trigger, it is hard for slices to adjust the resources in time due to resource competition among physical/virtual nodes. Besides, the resource provisioning strategies can also be optimized when slices are running normally, which is seldom considered when performing slice anomaly detection. To cope with these challenges, in this paper, we are motivated to locate the potential slice anomalies and optimize the resource allocation strategies in a holistic view by learning users' historical behaviors. Specifically, we design a general network architecture, model the process of slice resource provisioning, and formulate the problem as maximizing the long-term system net promoter score (NPS). To solve this problem, we propose a framework to locate the potential slice anomalies and decide the resource allocation strategies simultaneously by predicting the users' future requests and positions. As a result, simulation results demonstrate that our proposed scheme outperforms other baselines in improving the long-term system NPS and reducing the average latency of users.

Human-centric industrial processes, such as logistics, inspection, maintenance, and complex assembly, heavily rely on human expertise and judgment. In today’s dynamic and complex manufacturing environments, enhancing operator perception is crucial for timely and accurate decision-making. To facilitate effective communication between human workers and the complex factory ecosystem, this research proposes a system framework leveraging Digital Twin (DT) and semantic technologies to manage industrial heterogeneous data and provide operators with real-time insights. The system architecture comprises three primary layers: the Field Layer, the Information and Service Layer, and the Application Layer. The Information Layer integrates four core engines: Knowledge Engine for managing process-specific knowledge, Data Engine for handling streaming data, Artificial Intelligence (AI) Engine for incorporating advanced machine learning models, and 3D Engine for virtual representation and simulation. This paper presents a detailed implementation of the proposed system framework and validates it through a practical in-plant logistics transport use case. Results demonstrate the framework’s effectiveness in enhancing operator perception and decision-making by providing intuitive interfaces and timely insights.

The integration of communication and radar systems could enhance the robustness of future communication systems to support advanced application demands, e.g., target sensing, data exchange, and parallel computation. In this article, we investigate the beamforming design for integrated sensing, computing, and communication (ISCC) in the Internet of Robotic Things (IoRT) scenario. Specifically, we assume that each robot uploads its preprocessed sensing information to the access point (AP). Meanwhile, leveraging the additive features of the spatial wireless channels between robots and AP, over-the-air computation (AirComp) through multirobot cooperation could bolster system performance, particularly in tasks like target localization through sensing. To get a full picture of the effects of antenna array structures and beampatterns on the ISCC system, we evaluate the performance by considering the shared and separated antenna structures, as well as the omnidirectional and directional beampatterns. Based on these setups, the nonconvex optimization problems for the performance tradeoff between sensing and AirComp are formulated to minimize the mean-squared error (MSE) of AirComp and sensing. To efficiently solve these optimization problems, we designed the gradient descent augmented Lagrangian (GDAL) algorithm, which involves dynamically adjusting the step sizes while updating the variables. Simulation results show that the separated antenna structure achieves a lower AirComp MSE than the shared antenna setup because it has greater beam steering Degrees of Freedom. Moreover, the beampattern types have almost no effect on the AirComp MSE for the given antenna structure setup. This comprehensive investigation provides useful guidelines for ISCC framework implementation in IoRT applications.

Multiaccess edge computing (MEC) provides services for resource-sensitive and delay-sensitive Internet of Things (IoT) applications by extending the capabilities of cloud computing to the edge of the networks. However, the high mobility of IoT devices (e.g., vehicles) and the limited resources of edge servers (ESs) affect the service continuity and access latency. Service migration and reasonable resource (re-)allocation consequently become needed to ensure Quality of Service (QoS). However, service migration results in additional latency. In addition, different mobile IoT users have different resource requirements and different resource allocation policies of target ESs also determine whether service migration is necessary. Subsequently, how to jointly optimize service migration and resource allocation (SMRA) is a challenge that needs to be carefully addressed. To this end, this article investigates the joint optimization problem of SMRA in MEC environments to minimize the access delay of IoT users. It proposes a joint SMRA algorithm based on deep reinforcement learning (DRL), which takes into account the mobility of IoT users and decides whether to migrate services, where to migrate, and how to allocate resources through the long short time memory (LSTM) algorithm and the parameterized deep$Q$-network (PDQN) algorithm. Moreover, the PDQN algorithm effectively solves the discrete-continuous hybrid action space challenge in the SMRA problem. Finally, we conduct evaluation using a real-world data set of Beijing cab trajectories to verify the effectiveness and superiority of our proposed SMRA solution.

Underwater acoustic sensor networks (UASNs) have been widely deployed in many areas, such as marine ranching, naval applications, and marine disaster warning systems. The security of UASNs, particularly insider threats, is of growing concern. Internal attacks carried out via compromised normal nodes are more damaging and stealthy than external attacks, such as signal stealing, data decryption, and identity forgery. As a security mechanism for internal threat detection based on interaction data, trust models have proven to enhance the security of UASNs. However, traditional trust models lack sufficient scalability when faced with movable underwater devices, heterogeneous network environments, and variable attack patterns. Therefore, in this paper, a novel trust model based on federated deep reinforcement learning is proposed for UASNs. First, the evidence acquisition mechanism, including communication, energy, and data evidence, is improved based on existing ones to better accommodate the topological dynamics of UASNs. Second, acquired trust evidence is fed into the corresponding deep reinforcement learning-based local trust model to accomplish trust prediction and model training. Finally, a federated learning-based update method periodically aggregates and updates the parameters of the local models. The experimental results prove that the proposed scheme exhibits satisfactory performance in terms of improving trust prediction accuracy and energy efficiency.

Network slices are generally coupled with services and face service continuity/unavailability concerns due to the high mobility and dynamic requests from users. Network slice mobility (NSM), which considers user mobility, service migration, and resource allocation from a holistic view, is witnessed as a key technology in enabling network slices to respond quickly to service degradation. Existing studies on NSM either ignored the trigger detection before NSM decision-making or didn't consider the prediction of future system information to improve the NSM performance, and the training of deep reinforcement learning (DRL) agents also faces challenges with incomplete observations. To cope with these challenges, we consider that network slices migrate periodically and utilize the prediction of system information to assist NSM decision-making. The periodical NSM problem is further transformed into a Markov decision process, and we creatively propose a prediction-based federated DRL framework to solve it. Particularly, the learning processes of the prediction model and DRL agents are performed in a federated learning paradigm. Based on extensive experiments, simulation results demonstrate that the proposed scheme outperforms the considered baseline schemes in improving long-term profit, reducing communication overhead, and saving transmission time.

The popularity of microservices in industry has sparked much attention in the research community. Despite significant progress in microservice deployment for resource-intensive services and applications at the network edge, the intricate dependencies among microservices are often overlooked, and some studies underestimate the importance of system context extraction in deployment strategies. This paper addresses these issues by formulating the microservice deployment problem as a max-min problem, considering system cost and quality of service (QoS) jointly. We first study the attention-based microservice representation (AMR) method to achieve effective system context extraction. In this way, the contributions of different computing power providers (users, edge servers, or cloud servers) in the networks can be effectively paid attention to. Subsequently, we propose the attention-modified soft actor-critic (ASAC) algorithm to tackle the microservice deployment problem. ASAC leverages attention mechanisms to enhance decision-making and adapt to changing system dynamics. Our simulation results demonstrate ASAC's effectiveness, prioritizing average system cost and reward compared to the other state-of-the-art algorithms.

Beyond 5G applications, future 6G services would need to support very large data volumes for emerging industry verticals, such as holographic-type communications, as well as time-sensitive services, e.g., industrial control. Network slicing is the key technology to deliver such customizable services. Slices and their dedicated resources should be provisioned optimally where the services will be run with low network latencies and associated expenses. However, the user dynamics on resource demands within and between slices result in different resource re-allocation triggers, ultimately lead to distinct mobility patterns, e.g., scaling, migration, where sufficient resources must be transferred. Efficient slice mobility requires increasing flexibility in network operation and management to ensure the customized QoS while minimizing the corresponding mobility cost. In this paper, a prediction-based intelligent network analytic is proposed to facilitate the optimized network slice mobility scheme. We will investigate how to utilize the user and network prediction as the auxiliary information to make the slice mobility decision with the objective of maximizing the long-term profits while minimizing the latency and mobility cost. Finally, we evaluate the proposed prediction-based network slice mobility scheme in a simulated environment and compare its performance in terms of system costs, revenues, and profits with two benchmark solutions.

Deterministic networking (DetNet) has recently drawn much attention by investigating deterministic flow scheduling. Combined with artificial intelligent (AI) technologies, it can be leveraged as a promising network technology for facilitating automated network configuration in the Industrial Internet of Things (IIoT). However, the stricter requirements of the IIoT have posed significant challenges, that is, deterministic and bounded latency for time-critical applications. This paper incorporates deep reinforcement learning (DRL) in Cycle Specified Queuing and Forwarding (CSQF) and proposes a DRL-based Deterministic Flow Scheduler (Deep-DFS) to solve the Deterministic Flow Routing and Scheduling (DFRS) problem. Novel delay aware network representations, action masking and criticality aware reward function design are proposed to make Deep-DFS more scalable and efficient. Simulation experiments are conducted to evaluate the performances of Deep-DFS, and the results show that Deep-DFS can schedule more flows than the other benchmark methods (heuristic-based and AI-based methods).

With the increasing demand of network slices in vertical industries, slice resource provisioning in transport networks has encountered two challenges, one is efficient slice resource provisioning in the presence of traffic uncertainty of slices, and another is flexible slice resource isolation for customizable isolation needs. In this paper, we propose an innovative flexible hybrid isolation model to support any customized resource isolation from complete isolation to full sharing, and solve the slice resource provisioning problem named Hybrid Slicing Minimum Bandwidth (HSMB) by considering traffic prediction error to mitigate the negative impact of traffic uncertainty in the proposed model. After analyzing the HSMB problem, 1) we first try to solve the problem in steps and decompose the HSMB problem into grouping sub-problem and adjusting sub-problem, 2) we then propose a low-complexity dynamic programming grouping algorithm and a fast iterative adjustment algorithm for the two sub-problems based on probabilistic feature-based analysis, 3) we combine the algorithms of the two sub-problems and further propose a linking algorithm for the potential insufficient resource dilemma and high computational complexity dilemma to improve the efficiency of the solution. The numerical results show that the proposed flexible hybrid isolation model with different factors can facilitate flexible slice isolation with customized isolation demands, while the proposed algorithm can realize efficient slice resource provisioning with a probabilistic guarantee. The comparison result shows the proposed algorithms outperform the other benchmark algorithms.

Recently, we have observed an explosion in the intellectual capacity of user equipment, coupled by a meteoric rise in the need for very demanding services and applications. The majority of the work leverages edge computing technologies to accomplish the quick deployment of microservices, but disregards their inter-dependencies. In addition, while constructing the microservice deployment approach, several research disregard the significance of system context extraction. The microservice deployment issue (MSD) is stated as a max-min problem by concurrently evaluating the system cost and service quality. This research first analyzes an attention-based microservice representation approach for extracting system context. The attention-modified soft actor-critic method is proposed to the MSD issue. The simulation results reveal the ASAC algorithm's priorities in terms of average system cost and system reward.

Deterministic Networking (DetNet) has recently attracted much attention. It aims at studying the deterministic bounded latency and low latency variation for time-sensitive applications (e.g., industrial automation). To improve the quality of service (QoS) guarantee and make the network management efficient, it is desirable for Internet Service Provider (ISP) to obtain an optimal service function chain (SFC) provision strategy while providing deterministic service performance for the time-sensitive applications. In this paper, we will study the deterministic SFC lifetime management problem in beyond 5G edge fabric with the objective of maximizing the overall profits and ensuring the deterministic latency and jitter of SFC requests. We first formulate this problem as a mathematical model with the maximal profits for ISP. Then, the novel Deterministic SFC Deployment algorithm (Det-SFCD) and SFC Adjustment algorithm (Det-SFCA) due to traffic load variation are proposed to efficiently solve the SFC lifetime management problem. Extensive simulation results show that our proposed algorithms can achieve better performance in terms of SFC request acceptance rates, overall profits and latency variation compared with the benchmark algorithm.

Along with the increasing demand for latency-sensitive services and applications, Deterministic Network (DetNet) concept has been recently proposed to investigate deterministic latency assurance for services featured with bounded latency requirements in 5G edge networks. The Network Function Virtualization (NFV) technology enables Internet Service Providers (ISPs) to flexibly place Virtual Network Functions (VNFs) achieving performance and cost benefits. Then, Service Function Chains (SFC) are formed by steering traffic through a series of VNF instances in a predefined order. Moreover, the required network resources and placement of VNF instances along SFC should be optimized to meet the deterministic latency requirements. Therefore, it is significant for ISPs to determine an optimal SFC deployment strategy to ensure network performance while improving the network revenue. In this paper, we jointly investigate the resource allocation and SFC placement in 5G edge networks for deterministic latency assurance. We formulate this problem as a mathematic programming model with the objective of maximizing the overall network profit for ISP. Furthermore, a novel Deterministic SFC deployment (Det-SFCD) algorithm is proposed to efficiently embed SFC requests with deterministic latency assurance. The performance evaluation results show that the proposed algorithm can provide better performance in terms of SFC request acceptance rate, network cost reduction, and network resource efficiency compared with benchmark strategy.