标题： 上行小小区网络中的高效能计算卸载与虚拟连接控制 显示英文标题

标题： Energy Efficient Computation Offloading and Virtual Connection Control in Uplink Small Cell Networks

摘要： 如今，在机器学习的框架下，软计算技术在电力系统中的应用正在增加，并受到良好的接受。 在本文中，我们首先提出一种深度学习方法，以找到HetNet系统的最佳配置。 我们使用了一个测试系统的大数量径向配置来进行训练。 除了确保所有用户的体验质量外，我们还研究了多层分层网络中的联合载波/功率分配问题，以实现最优的功率效率。 所提出的方法使用了一个自适应负载平衡模型，旨在从关键性能指标的角度实现所有服务器之间的“几乎最优”公平性。 与当前基于模型的能效方法不同，我们提出了一种联合资源分配、能效和流量控制算法，以解决常见的非凸和分层优化问题。 此外，通过参考基于SLA的连续资源分配，我们将所提出的算法扩展到常见的流量/功率控制和运营功率优化算法，以在确保用户吞吐量限制的同时实现最优能效。 此外，仿真结果表明，所提出的受控功率/流量优化方法相比使用网络拓扑调整能力的传统设计可以显著提高能效。 显示英文摘要

摘要： Nowadays, the use of soft computational techniques in power systems under the umbrella of machine learning is increasing with good reception. In this paper, we first present a deep learning approach to find the optimal configuration for HetNet systems. We used a very large number of radial configurations of a test system for training purposes. We also studied the issue of joint carrier/power allocation in multilayer hierarchical networks, in addition to ensuring the quality of experience for all subscribers, to achieve optimal power efficiency. The proposed method uses an adaptive load equilibrium model that aims to achieve "almost optimal" equity among all servers from the standpoint of the key performance indicator. Unlike current model-based energy efficiency methods, we propose a joint resource allocation, energy efficiency, and flow control algorithm to solve common nonconvex and hierarchical optimization problems. Also, by referring to the allocation of continuous resources based on SLA, we extended the proposed algorithm to common flow/power control and operational power optimization algorithm to achieve optimal energy efficiency along with ensuring user's throughput limitations. Also, simulation results show that the proposed controlled power/flow optimization approach can significantly increase energy efficiency compared to conventional designs using network topology adjustment capability.