Quality of Service-Adaptive Industrial Internet of Things leveraging Edge Computing and Deep Reinforcement Learning : The Deep QoS-Adaptive Learning Environment (DeQALE) Architecture

Abstract

The advent of the Internet of Things (IoT) paradigm allows for real-world tasks to be monitorized and managed using computing applications. The application of IoT to the industrial environment leads to the Industrial Internet of Things (IIoT), in which industrial processes are managed through IoT, thus allowing industry workers to better control their facilities and processes. However, IIoT applications have very strict Quality of Service (QoS) requirements, such as short response times, that require for the deployment of their services in edge nodes, close to the facilities. In IIoT scenarios, deploying each of the services so that the QoS requirements are met is not an easy task. Moreover, the dynamicity of the environment requires for a fast, adaptive solution. In this position paper, we propose DeQALE, an approach to train a deep reinforcement learning agent to solve this problem in short cycles.