Auto scheduling through distributed reinforcement learning in SDN based IoT environment

Abstract

Abstract The Internet of Things (IoT), which is built on software-defined networking (SDN), employs a paradigm known as channel reassignment. This paradigm has great potential for enhancing the communication capabilities of the network. The traffic loads may be scheduled more effectively with the help of an SDN controller, which allows for the transaction of matching channels via a single connection. The present techniques of channel reassignment, on the other hand, are plagued by problems with optimisation and cooperative multi-channel reassignment, which affect both traffic and routers. In this paper, we provide a framework for SDN–IoT in the cloud that permits multi-channel reassignment and traffic management simultaneously. The multi-channel reassignment based on traffic management is optimised via the use of a deep reinforcement learning technique, which was developed in this paper. We do an analysis of the performance metrics in order to optimise the throughput while simultaneously reducing the rate of packet loss and the amount of delay in the process. This is achieved by distributing the required traffic loads over the linked channels that make up a single connection.

Keywords

• Software-defined networking
• Internet of Things
• Channel assignment
• Traffic management
• Reinforcement learning
• Multi-channel reassignment