Routing optimization in cloud-edge collaborative networks faces a fundamental conflict between global strategic planning and local real-time responsiveness, further complicated by structural heterogeneity and stochastic traffic patterns. Traditional protocols lack adaptivity, while existing Deep Rein-forcement Learning (DRL) approaches based on Graph Neural Networks (GNN) struggle with limited receptive fields and over-smoothing issues in large-scale topologies. In this paper, we propose HAT-Route, a Transformer-driven hierarchical routing framework supported by the Network Digital Twin (NDT). Our contributions are threefold: (1) We establish a cloud-edge collaborative architecture operating under the Centralized Training and Decentralized Execution paradigm. This architecture balances the trade-off between global optimization and real-time inference. (2) We introduce FlowFormer, a Spatiotemporal Transformer for the NDT. FlowFormer integrates a novel Edge-Conditioned Spatial Attention (EC-SAT) mechanism to capture physical link constraints and distinguish between congestion and Head-of-Line (HOL) blocking. (3) We design HAT-Route, a hierarchical DRL agent that utilizes Graph Transformers for global policy learning in the cloud, coupled with knowledge distillation to deploy lightweight policies at the network edge. Extensive experiments demonstrate that our framework outperforms traditional protocols and GNN-based baselines in terms of QoS optimization, training stability, scalability, and generalization capability on large-scale network topologies.

Emerging industrial Internet-of-Things (IoT) applications demand diverse and critical Quality of Service (QoS). Deep reinforcement learning (DRL)-based routing approaches offer promise but struggle with scalability and convergence, particularly when dealing with graph-based network information. To tackle the challenge, we propose a distributed routing model that leverages graph representation learning (GRL) to learn the optimal routing decision in a distributed manner. We further present on-demand routing algorithms composed of graph representation learning (GRL)-based feature engineering and DRL-based routing decision-making to meet differential QoS requirements. Experimental results demonstrate our approach outperforms state-of-the-art DRL-based routing algorithms in a distributed manner, particularly in large-scale and heavy-load networks.

Time-critical applications, such as remote surgery, virtual reality, and automatic driving, impose stringent requirements for minimizing delay and jitter. The queuing delay at forwarding nodes significantly affects application traffic. To tackle this challenge, Time-Sensitive Networking (TSN) techniques have been proposed to ensure deterministic queuing delay. However, existing deterministic solutions, including the Time-Aware Shaper (TAS) and the Cyclic Queuing and Forwarding (CQF) model, encounter limitations due to their reliance on precise synchronization and dedicated forwarding plane devices. In this paper, we present a novel solution: the Strict Priority queuing model with a queuing Delay Guarantees (SP-DG), specifically designed for programmable forwarding planes. This model introduces an upper bound on the queuing delay for packets at each switch, effectively characterizing their distinct queuing delay requirements. By utilizing exponential smoothing to predict traffic intensity, packets are intelligently scheduled to appropriate queues based on their upper-bound queuing delay requirements. Our Netbench simulation and P4 software switch experiments demonstrate that the proposed SP-DG algorithm, compared to the state-of-the-art SP-based queue scheduling algorithms, achieves superior performance in delay-bounded queue scheduling.

As short video applications gain popularity, researchers are exploring ways to enhance Quality of Experience (QoE) for short videos while maximizing network bandwidth efficiency. Despite the growing interest, existing Adaptive Bitrate (ABR) algorithms primarily concentrate on content prefetching strategies and often overlook the dynamic interaction between network congestion control and ABR. This interaction is especially critical for short video streaming, where network conditions can fluctuate rapidly, and user expectations for seamless playback are high. To address these challenges, we propose aCroSS, an AI-driven framework for adaptive short video streaming that jointly optimizes both the application and transport layers to enhance QoE and bandwidth utilization. The aCroSS algorithm leverages advanced machine learning techniques to adapt in real time to fluctuating network conditions and dynamic user behaviors, delivering a more robust and responsive streaming experience. Our simulation results demonstrate that aCroSS consistently outperforms existing baseline algorithms, achieving more than a 10% improvement in utility scores across various network trace datasets. This highlights the effectiveness of aCroSS in delivering superior performance in diverse streaming environments.

In upcoming sixth generation (6G) networks, it is a critical challenge to support a plethora of innovative services across wide-area networks. To realize the dedicated QoS provisioning and meet the diverse quality of service (QoS) requirements of services in terms of criteria like bandwidth, delay, jitter and loss ratio, we proposed an AI-driven on-demand routing framework to support the diverse end-to-end QoS Provisioning in large-scale wide-area networks. Specifically, we make further efforts on solving the instability and non-convergence issues of the AI-driven routing algorithm and enhance it with the assistance of expert knowledge on traffic engineering and the latest advance on reinforcement learning. Furthermore, the simulation results show that our algorithm outperforms other benchmark routing algorithms with efficient learning and a significant reduction in delay, jitter and loss ratio by the traffic data sets of the real-world wide-area networks.

With the rapid development of Internet applications, diversified Quality of Service (QoS) has been required in packet routing to meet the demand of various types of applications. In this article, an Intelligent QoS on-demand Routing (IQoR) framework has been proposed to support multiclass QoS Provisioning for packet forwarding. In addition, we present an IQoR with Link State Estimation (IQoR-LSE) algorithm with the assistance of link congestion inference to guide the exploration of action space in deep reinforcement learning, to seek optimal routing policies. The IQoR-LSE algorithm is proposed to solve the non-convergence problem at exploring the high-dimensional action space by jointly estimating the link congestion. Extensive simulations show that IQoR-LSE algorithm outperforms other benchmark routing algorithms with efficient learning and a significant reduction in average delay, jitter and packet loss.

Deterministic network needs to ensure the deterministic transmission requirements of different applications in terms of delay, packet loss rate, jitter, throughput, and reliability.In response to the differentiated and deterministic network transmission requirements of applications, an on-demand intelligent routing framework OdR for deterministic network was proposed.Under the OdR framework, an on-demand intelligent routing algorithm named OdR-TD3 based on deep reinforcement learning was proposed, which generates routing strategies based on the deterministic QoS requirements of application traffic, to satisfy the applications’ requirements of deterministic network.The experimental evaluation results show the OdR-TD3 algorithm has a significant advantage over the DV algorithm and the SPF algorithm in terms of the achievement rate of deterministic QoS requirements.

Augmented vehicular reality (AVR) is one of the key technologies to realize intelligent transportation in the future, which can significantly improve traffic safety and transportation efficiency of autonomous driving. However, available computation and spectrum resources of vehicles are not well utilized to meet the requirements of cooperative perception of autonomous vehicles. In order to meet the needs of sharing sensing data to cooperatively perceive the surrounding environment, executing delay-sensitive and computationally intensive tasks of autonomous driving applications, we propose computation offloading and resource allocation optimization for AVR (CoroAVR) algorithm with hybrid sensing data fusion of cooperative perception to process the real-time data in fog-edge computing for maximizing system throughput and spectrum utilization on the premise of ensuring the quality of task completion. First, the minimum signal to interference plus noise ratio needed to complete the task is obtained with the given maximum computing resources. Second, the optimal power allocation is carried out by using convex optimization theory, and the throughput gain is calculated, and the offloading decision is made by comparing the throughput gain. Finally, the channel allocation problem is solved by using the maximum matching algorithm of the bipartite graph, and the computation resource allocation is studied. The simulation results show that the performance of the proposed algorithm is better than that of the contrast algorithm.

The deterministic network is oriented to various services in the network, ensuring the deterministic requirements of different services in terms of delay, packet loss rate, cost overhead, throughput, security, and reliability. This paper proposes an on-demand routing learning model with a three-layer logical plane that can generate on-demand routing strategies with deterministic QoS (Quality of Services) requirements for different traffic flows. With the help of deep reinforcement learning algorithms, the OdR-TD3 and OdR-SAC algorithms are proposed to generate the suitable routing strategies to meet the deterministic QoS requirements of applications' traffic. The experimental evaluation results show the OdR-TD3 and OdR-SAC algorithm have significant advantages over the DV and SPF algorithms, considering the achievement rate of deterministic QoS requirements.

The future network is expected to support extremely large bandwidth, ultra-low latency or deterministic delay, extremely high reliability, and massive connectivity for novel forward-looking scenarios. As one of the most fundamental parts of the network, routing plays a vital role in a well-performed network. Recently, some new techniques using machine intelligence to optimize the network routing have been proposed. Although they have demonstrated great potential to improve the network performance, it is still a great challenge to apply machine intelligence-based routing in real network environment due to the limitations of current network architectures and protocols. Fortunately, the future network research program is on-going for designing new network paradigms, which provides an opportunity to address those limitations. In this paper, we investigate state-of-the-art techniques in machine intelligence-enabled network routing and discuss the development trends of machine intelligence-enabled routing optimization techniques for the future network.

With the rapid development of Internet applications, diversified Quality of Service (QoS) has been required in packet routing to meet the demand of various types of applications. This paper presents an Intelligent QoS-aware Routing (IQoR) framework with the assistance of Deep Reinforcement Learning (DRL), which supports multi-class QoS provisioning for packet forwarding. The simulation results show that IQoR outperforms the widely-used benchmark routing algorithms by significantly reducing the average delay and jitter of packets.

The radical suppression of the photodarkening effect and laser performance deterioration via H 2 loading were demonstrated in high-power Yb-doped fiber（YDF） amplifiers. The photodarkening loss at equilibrium was114.4 d B/m at 702 nm in the pristine fiber, while it vanished in the H 2 -loaded fiber. To obtain a deeper understanding of the impact of photodarkening on laser properties, the evolution of the mode instability threshold and output power in fiber amplifiers was investigated. After pumping for 300 min, the mode instability threshold of the pristine fiber dropped from 770 to 612 W, and the periodic fluctuation of the output power became intense,finally reaching 100 W. To address the detrimental effects originating from photodarkening, H 2 loading was applied in contrast experiments. The output power remained stable, and no sign of mode instability was observed in the H 2 -loaded fiber. Moreover, the transmittance at 638 nm confirmed the absence of the photodarkening effect. The results pave the way for the further development of high-power fiber lasers.

Faced with the severe challenges of today’s network in terms of scalability,mobility,QoS guarantee and manageability,the essential capability requirements of future networks were summarized by analyzing application scenarios of future networks,and the key functional requirements and performance indicators of future networks were proposed.Finally,the development trend of the future networks was forecasted.

A Mach–Zehnder interferometer (MZI) based on an etched all-solid microstructure fiber (MOF) has been demonstrated. The MZI works on the basis of interference between the vibrant core and cladding modes in the MOF. The all-solid MOF has a heterostructure cladding composed of Ge-doped rod arrays and pure silica, and thus can support and propagate a vibrant cladding mode with a large mode area. When the outermost cladding of MOF is etched, the cladding mode becomes sensitive to the ambient refractive index (RI). The etched MOF can work as a sensing head for RI sensing. By comparing the interference spectra, the extinction ratio has remained stable at around 20 dB after the MOF was etched. The RI sensing characteristics of the MZI with an etched MOF have also been investigated. The results show that the RI sensitivity can reach up to 2183.6 nm/RIU with a low-temperature coefficient (<10 pm/°C).

Information theory progression along with the advancements being made in the field of Vehicular Ad hoc NETworks (VANETs) supports the use of coding-aware opportunistic routing for efficient data forwarding. In this work, we propose and investigate an adaptive coding-aware routing scheme in a specific VANET scenario known as a vehicular platoon. Availability of coding opportunities may vary with time and therefore, the accurate identification of available coding opportunities at a specific time is a quite challenging task in the highly dynamic scenario of VANETs. In the proposed approach, while estimating the topology of the network at any time instance, a forwarding vehicle contemplates the composition of multiple unicast data flows to encode the correct data packets that can be decoded successfully at destinations. The results obtained by using OMNeT++ simulator reveal that higher throughput can be achieved with minimum possible packet transmissions through the proposed adaptive coding-aware routing approach. In addition, the proposed adaptive scheme outperforms static transmissions of the encoded packets in terms of coding gain, transmission percentage, and encoded packet transmission. To the best of our knowledge, the use of coding-aware opportunistic routing has not been exploited extensively in available literature with reference to its implications in VANETs.

We demonstrated a kind of long-period fiber grating (LPFG), which is manufactured with a thermal diffusion treatment. The LPFG was inscribed on an ultrahigh-numerical-aperture (UHNA) fiber, highly doped with Ge and P, which was able to easily diffuse at high temperatures within a few seconds. We analyzed how the elements diffused at a high temperature over 1300 °C in the UHNA fiber. Then we developed a periodically heated technology with a CO2 laser, which was able to cause the diffusion of the elements to constitute the modulations of an LPFG. With this technology, there is little damage to the outer structure of the fiber, which is different from the traditional LPFG, as it is periodically tapered. Since the LPFG itself was manufactured under high temperature, it can withstand higher temperatures than traditional LPFGs. Furthermore, the LPFG presents a higher sensitivity to high temperature due to the large amount of Ge doping, which is approximately 100 pm/°C. In addition, the LPFG shows insensitivity to the changing of the environment’s refractive index and strain.

Applying Application-Level Multicast technology (ALM) to Distributed Interactive Applications (DIAs) becomes more and more popular. Especially for DIAs embedded priority. The PST algorithm was designed for these DIAs based on distance. However, the PST lacks efficient priority selection and filtering mechanism resulted in system unstable and inextensible. In this paper, we first propose PQPST algorithm which can predict every efficient priority and quantize the predicted efficient priorities into different groups for constructing the multicast trees. Second, we propose Priority Discrepancy Heuristic Mechanism (PDHM), which sets different thresholds of priority discrepancy within the priority discrepancy interval to control the distribution tree construction, and it can efficiently decrease the repeated distribution tree construction. According to the simulation results, the PQPST and PDHM can efficiently improve the performance of the PST algorithm.

With the rapid growth of cloud applications and users, Data Center Network (DCN) has become a critical component in the cloud ecosystem to sustain remarkable computation demand. It is facing great challenges in performance guarantee, security enforcement, and resource and energy management. Software Defined Networking (SDN), as an efficient way to collect network information and perform network management, has been adopted in DCN to enable automated network configuration and management. In this paper we aim at surveying the state-of-the-art techniques of using SDN in DCN and discuss the way SDN help DCN. We first present the introduction of SDN and DCN, then survey the SDN based DCN, and finally show some lessons and future trend.

Software Defined Networking (SDN) is a revolutionary network architecture that separates out network control functions from the underlying equipment and is an increasingly trend to help enterprises build more manageable data centers where big data processing emerges as an important part of applications. To concurrently process large-scale data, MapReduce with an open source implementation named Hadoop is proposed. In practical Hadoop systems one kind of issue that vitally impacts the overall performance is know as the NP-complete minimum make span problem. One main solution is to assign tasks on data local nodes to avoid link occupation since network bandwidth is a scarce resource. Many methodologies for enhancing data locality are proposed such as the HDS and state-of-the-art scheduler BAR. However, all of them either ignore allocating tasks in a global view or disregard available bandwidth as the basis for scheduling. In this paper we propose a heuristic bandwidth-aware task scheduler BASS to combine Hadoop with SDN. It is not only able to guarantee data locality in a global view but also can efficiently assign tasks in an optimized way. Both examples and experiments demonstrate that BASS has the best performance in terms of job completion time. To our knowledge, BASS is the first to exploit talent of SDN for big data processing and we believe it points out a new trend for large-scale data processing.