Machine Learning for Caching Placement in Edge Computing Networks

Internet of Things devices (IoTDs) that deployed for various applications have limited size, power, storage, and computing capabilities, and many of them are used for low-latency applications. Mobile edge computing (MEC) can be leveraged to reduce the latency of IoTDs, and judicious caching placement can help further to improve the quality of service. In this paper, we study multi-content placement (MCP) in the MEC networks for IoTDs by considering the background data caching in the edge nodes and the data collection of IoTDs. The MCP problem is formulated by considering the caching, the IoTD assignment, and the communication and computing resources allocation to minimize the average latency of all IoTDs. As the MCP problem is NP-hard, we propose a deep reinforcement machine learning algorithm to obtain the caching placement and IoTD assignment jointly to solve the MCP problem. We use an optimal joint resource-scheduling algorithm to assign the resources. Our results demonstrate that considerable latency improvement can be achieved through the proposed deep reinforcement machine learning algorithm as compared to baseline algorithms.