This work proposes Double-Channel Graph Attention (DCGA), an end-to-end reinforcement learning framework that isolates network reachability and demand-service logic into separate graph channels and constructs valid routes using a simulator-coupled, constraint-informed decoder.
Abstract
We study an integrated pickup-and-delivery problem on sparse, non-Euclidean networks that jointly optimizes cyclic routing, cargo flow allocation, and cross-cycle service. The tight coupling of these operational constraints creates a complex discrete-continuous decision space with highly restricted feasible regions. To overcome these computational challenges, we propose Double-Channel Graph Attention (DCGA), an end-to-end reinforcement learning framework. DCGA isolates network reachability and demand-service logic into separate graph channels and constructs valid routes using a simulator-coupled, constraint-informed decoder. Experiments on LinerLib benchmarks demonstrate that DCGA achieves seconds-level inference and delivers state-of-the-art solution quality on instances beyond a specific scale, with its advantage over existing baselines widening significantly as problem size increases. Supported by extensive stability and ablation analyses, our results demonstrate that this structure-aware learning approach provides an effective, low-latency engine for realistic routing-and-flow optimization.
Networks with highly dynamic data transmission demands and network topologies are common in real world. A fundamental problem in such networks is achieving scalable traffic allocation to maximize long-term total throughput under link capacity constraints. However, state-of-the-art (SOTA) works lack scalability. This is primarily due to two reasons in large-scale networks: first, they require solving constrained optimization problems online, which leads to high decision latency; second, they rely on reinforcement learning algorithms for policy optimization, which are inefficient in exploration and challenging to train effectively. To address these issues, we propose the Fast Networked Control (FNC) policy framework, which firstly utilizes parallelizable neural network modules to process the state and generate raw decisions, followed by basic operations such as normalizations and comparisons, which do not require iteration or optimization, to obtain decisions that satisfy the constraints. Hence, FNC policy avoids solving constrained optimization problems and supports parallel execution, significantly reducing decision latency. Furthermore, this policy preserves gradient flow and supports backpropagation, which enable us to design an imitation learning algorithm to efficiently train the policy in an end-to-end manner. Experiments in large-scale networks show that our FNC policy achieves an average 8% improvement in demands satisfaction and 10 times reduction in decision latency versus SOTA works.
Zhaoxing Yang, Guiyun Fan, Anjie Cao et al.· IEEE Transactions on Network...· 0 citations
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 Reinforcement 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.
Bin Dai, Yuntao Wang, Jianhai Zheng· IEEE Transactions on Network...· 0 citations
Fast and stable routing convergence is critical in large IP networks, and the interior-gateway-protocol (IGP) timers that govern failure detection (Hello/Dead intervals) expose a fundamental tension: aggressive timers detect failures quickly but inflate control overhead and trigger route flaps, whereas conservative timers are stable but slow. Recent work such as DRL-Adapt has shown that deep reinforcement learning can tune these timers better than static defaults, but it operates on a flat, globally-aggregated network state and emits a single network-wide timer, so it can neither exploit the spatial heterogeneity of real topologies nor transfer architecturally across networks of different size. We propose GraphRoute-Transfer, a graph-neural-network policy that assigns per-node timers from local structural features and is by construction permutation- and size-invariant. Because control-plane fragility and failure criticality are spatially heterogeneous, the cost-minimizing timer assignment varies across the graph; our policy learns this mapping and applies it zero-shot to unseen topologies of arbitrary size. Training is guided by a coordinate-descent search oracle on a convergence-cost objective, so the expensive per-topology optimization is amortized into a sub-millisecond inference. On 231 real topologies from the Internet Topology Zoo, GraphRoute-Transfer reduces mean convergence time by 37.3% relative to the OSPF default and to a flat DRL baseline, lowers the composite convergence-cost objective by 16.4% over the flat baseline, and attains 1.317 cost—within 0.3% of the search oracle—while running about 8,160× faster than the search. Crucially, a policy trained only on networks with ≤70 nodes maintains its gains on unseen networks up to 140 nodes, whereas the flat baseline degenerates to a global constant that cannot adapt.
Yuto Nakamura· Journal of Computing and Ele...· 0 citations
Space–Air–Ground Integrated Networks (SAGINs) have been envisioned to support next-generation communication networks.Due to their heterogeneity and dynamic link characteristics, routing in SAGIN is challenging. Existing shortest-path routing mechanisms do not adapt well to the varying bandwidth and latency, leading to poor quality of service (QoS). In this paper, we propose a deep reinforcement learning (DRL)-based adaptive routing scheme for maximizing throughput and minimizing end-to-end delay jointly in SAGIN. In the proposed model, an agent learns the policy of choosing the suitable path by interacting with the network environment and obtaining rewards. The network is modeled as a weighted graph with delay and bandwidth constraints. We compare our model with a traditional delay minimization baseline over multiple independent runs. Experimental results show that our DRL approach achieves a 6.51% improvement in average throughput and a 29.90% reduction in end-to-end delay compared to the baseline strategy. Statistical analysis confirms the robustness of the delay reduction, highlighting the effectiveness of reinforcement learning in dynamic HetNets. This indicates that adaptive policy learning enables better congestion avoidance and more efficient resource utilization. Overall, the proposed DRL-based routing framework offers a scalable and intelligent solution for optimizing performance in complex SAGIN architectures, with promising potential for next generation integrated communication systems.
As practical quantum networks approach large-scale deployment, the need for efficient user-to-user frequency allocation is increasing, yet current approaches only provide partial solutions to the routing and spectrum allocation problem for an arbitrary quantum network. We address this challenge for repeater-less flex-grid quantum networks based on hyperentangled photons using an efficient three-stage pipeline combining leading tools in classical networking with recent advances in numerical optimization. First, double instantiations of Yen's algorithm obtain low-loss route candidates between each pair of users and the entanglement sources. Second, the advanced process optimizer (APOPT) obtains frequency channel allocations that maximize distribution rates under fidelity constraints. Finally, the constraint programming solver using satisfiability methods (CP-SAT) assigns specific frequency bins to each link, ensuring that there is no contention between frequencies from different sources. We numerically demonstrate this approach on a representative ring network and a Manhattan incumbent local exchange carrier topology, realizing significant improvements over prior genetic algorithm approaches in speed, accuracy, and scalability. Overall, this pipeline provides an efficient heuristic workflow for optimizing broadband entanglement distribution, applicable to arbitrarily connected quantum networks integrated within the existing lightwave infrastructure.
Zachary Goisman, M. L. Stevens, Maxwell Goisman et al.· 0 citations
Networked systems, from power grids to traffic networks and cloud clusters, carry loads across nodes with limited capacity. A node whose load exceeds its capacity fails and sheds its load onto its neighbors, which can trigger a system-wide cascade. We study how to allocate a fixed capacity budget across nodes to resist these cascades under local load redistribution. The problem is difficult because no optimal allocation is known, and the fail-or-survive objective is non-differentiable and piecewise constant, so exact and gradient-based optimization methods do not directly apply. We introduce TANGCO (Topology-Aware Neural Graph-Guided Capacity Optimization), which uses a graph neural network policy trained through the cascade simulator with policy-gradient learning and a heuristic anchor. We evaluate TANGCO on five synthetic graph families and five real networks spanning power, road, air, and Internet topologies. The learned policy improves on the best of four hand-designed heuristics in all 450 synthetic instances and in 40 of 45 real-network conditions, with robustness gains ranging from 1.6% to 246%. The learned policies transfer to unseen graphs within a family and partially across related topologies, and TANGCO$^{pre}$, pre-trained on synthetic graphs, matches per-network training on unseen real networks. Training scales near-linearly with graph size, and TANGCO$^{pre}$ allocates on a new network with no per-target training, matching the deployment cost of a hand-designed heuristic. Free-vector variants without the GNN, stay close to the heuristics, so the graph representation carries the gain beyond numerical search. Finally, analysis of the learned allocations identifies when local risk is sufficient, leads to an improved closed-form heuristic, and reveals the regimes where a topology-aware learned policy remains necessary.