Multi-access edge computing (MEC) has attracted much more attention to revolutionising smart communication technologies and Internet of Everything. Nowadays, smart end-user devices are designed to execute sophisticated applications that demand more resources and explosively connected to the global ecosystem. As a result, the backhaul network traffic congestion grows enormously and user quality of experience is compromised as well. To address these challenges, the authors proposed congestion-aware adaptive decentralised computing, caching, and communication framework which can orchestrate the dynamic network environment based on deep reinforcement learning for MEC networks. MEC is a paradigm shift that transforms cloud services and capabilities platform at the edge of ubiquitous radio access networks in close proximity to mobile subscribers. The framework can evolve to perform augmented decision-making capabilities for the upcoming network generation. Hence, the problem is formulated using non-cooperative game theory which is nondeterministic polynomial (NP) hard to solve and the authors show that the game admits a Nash equilibrium. In addition, they have constructed a decentralised adaptive scheduling algorithm to leverage the utility of each smart end-user device. Therefore, their methodical observations using theoretical analysis and simulation results substantiate that the proposed algorithm can achieve ultra-low latency, enhanced storage capability, low energy consumption, and scalable than the baseline scheme.

中文翻译：

---

多访问边缘计算网络的拥塞感知自适应分散计算分流和缓存

多访问边缘计算（MEC）在革新智能通信技术和万物互联方面吸引了更多关注。如今，智能终端用户设备旨在执行需要更多资源并爆炸性地连接到全球生态系统的复杂应用程序。结果，回程网络流量拥塞急剧增加，并且用户体验质量也受到损害。为了应对这些挑战，作者提出了拥塞感知的自适应分散式计算，缓存和通信框架，该框架可基于针对MEC网络的深度强化学习来编排动态网络环境。MEC是一种范式转变，它转变了紧邻移动用户的无处不在的无线接入网络边缘的云服务和功能平台。该框架可以发展为未来的网络一代执行增强的决策能力。因此，该问题是使用非确定性多项式（NP）的非合作博弈理论来表述的 难以解决，作者证明游戏承认纳什均衡。此外，他们还构建了分散式自适应调度算法，以利用每个智能最终用户设备的效用。因此，他们使用理论分析和仿真结果进行的有条理的观察证实，与基线方案相比，该算法可实现超低延迟，增强的存储能力，低能耗和可扩展性。