Decentralized computation offloading and caching in Multi-Access Edge Computing (MEC) is a promising approach to evolve the forthcoming network generation. MEC is the emerging technology that provides adaptive micro cloud services to the edge of proximity resource-constrained smart communication and Internet of Everything (IoE) devices for cellular subscribers. Nowadays, Massive IoE devices are exponentially connected to the global ecosystem. As a result, the backhaul network traffic grows enormously and users’ ultra-reliable low latency communications are challenging as well. In this paper, we explored decentralized adaptive resource-aware communication, computing, & caching framework which can orchestrate the dynamic network environments based on Deep Reinforcement Learning (DRL). Subsequently, the framework can perform augmented decision-making capabilities to enhance users’ connectivity and resource utilization requirements. Basically, every IoE device user are attempting to capitalize their own utilities. Hence, the problem is formulated using Non-cooperative game theory which is non-deterministic polynomial to solve the structural property of the MEC networks. We analyze and show that the game admits a Nash Equilibrium. Moreover, we have introduced a decentralized cognitive scheduling algorithm by exploiting DRL technology to leverage the utility of IoE & smart communication devices. Therefore, numerical results and theoretical analysis revealed that the proposed algorithm outperform, ultra-reliable low latency, and scalable than the baseline schemes.

多访问边缘计算（MEC）中的分散式计算分载和缓存是发展即将到来的网络一代的一种有前途的方法。MEC是一种新兴技术，可为蜂窝用户提供自适应微云服务，以向邻近资源受限的智能通信和万物联网（IoE）设备的边缘提供服务。如今，大规模IoE设备已与全球生态系统指数级连接。结果，回程网络流量急剧增长，用户的超可靠低延迟通信也面临挑战。在本文中，我们探索了分散的自适应资源感知通信，计算和缓存框架，该框架可以基于深度强化学习（DRL）来编排动态网络环境。随后，该框架可以执行增强的决策功能，以增强用户的连接性和资源利用率要求。基本上，每个IoE设备用户都试图利用自己的实用程序。因此，该问题使用非确定性多项式非合作博弈论来解决，以解决MEC网络的结构特性。我们分析并表明，该游戏承认纳什均衡。此外，我们利用DRL技术引入了分散式认知调度算法，以利用IoE和智能通信设备的效用。因此，数值结果和理论分析表明，与基线方案相比，该算法具有更好的性能，超可靠的低等待时间和可扩展性。