Scholarlink

<p>Resource allocation in Self-Evolving Autonomous Networks (SEANs) remains a fundamental challenge due to the non-convex nature of decision functions, resource heterogeneity, and continuous topological dynamics. Traditional static or computationally intensive approaches often fail to deliver real-time efficiency under such dynamic conditions. To address these limitations, this paper introduces an OCO-SCA-based adaptive resource management framework that integrates Online Convex Optimization (OCO) with Successive Convex Approximation (SCA). The core concept relies on progressively transforming non-convex problems into locally convex subproblems, enabling tractable and efficient optimization. Through an eight-phase distributed decision-making process, the framework supports scalable, stable, and context-aware resource management by leveraging online learning and dynamic adjustment mechanisms. The proposed architecture consists of two complementary layers: (1) a learning and adaptation layer, which predicts future network behavior based on traffic history and resource variations, and (2) a decision and management layer, <i>whi</i>ch optimizes allocation policies and ensures network convergence through lightweight peer-to-peer information exchange. The framework’s performance was evaluated using three key metrics: Average Task Completion Time (ATCT), Resource Allocation Fairness Ratio (RAFR), and Average Topology Adaptation Time (ATAT). Simulation results demonstrate improvements of 16.26% in ATCT, 10.73% in RAFR, and 17.11% in ATAT compared to benchmark approaches. These findings confirm the proposed framework’s effectiveness in managing heterogeneous resources, ensuring fairness, and enhancing agility in SEANs, making it a practical solution for real-time applications in edge and fog environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Regret-aware adaptive resource allocation in self-evolving autonomous networks via OCO-SCA optimization

  • 期刊论文

  • 发表日期：

    2026年7月24日

查看全文
  • Ehsan Matinfar,
  • Marjan Mahmoudi,
  • Elham Maghami,
  • Behrang Barekatain,
  • Hamid R. Arabnia

摘要

Resource allocation in Self-Evolving Autonomous Networks (SEANs) remains a fundamental challenge due to the non-convex nature of decision functions, resource heterogeneity, and continuous topological dynamics. Traditional static or computationally intensive approaches often fail to deliver real-time efficiency under such dynamic conditions. To address these limitations, this paper introduces an OCO-SCA-based adaptive resource management framework that integrates Online Convex Optimization (OCO) with Successive Convex Approximation (SCA). The core concept relies on progressively transforming non-convex problems into locally convex subproblems, enabling tractable and efficient optimization. Through an eight-phase distributed decision-making process, the framework supports scalable, stable, and context-aware resource management by leveraging online learning and dynamic adjustment mechanisms. The proposed architecture consists of two complementary layers: (1) a learning and adaptation layer, which predicts future network behavior based on traffic history and resource variations, and (2) a decision and management layer, which optimizes allocation policies and ensures network convergence through lightweight peer-to-peer information exchange. The framework’s performance was evaluated using three key metrics: Average Task Completion Time (ATCT), Resource Allocation Fairness Ratio (RAFR), and Average Topology Adaptation Time (ATAT). Simulation results demonstrate improvements of 16.26% in ATCT, 10.73% in RAFR, and 17.11% in ATAT compared to benchmark approaches. These findings confirm the proposed framework’s effectiveness in managing heterogeneous resources, ensuring fairness, and enhancing agility in SEANs, making it a practical solution for real-time applications in edge and fog environments.