A Reinforcement Learning-Based Scheduling Scheme for the IEEE 802.15.4e TSCH Network

Abstract

Time-Slotted Channel Hopping (TSCH), as defined in the IEEE 802.15.4e amendment, is currently considered the most widely used Medium Access Control (MAC) protocol in Industrial Internet of Things (IIoT) wireless networks. However, the absence of a defined scheduling process within the standard remains an open area of research. This paper presents a proposal for integrating Q-Learning (QL), a Reinforcement Learning (RL) method, into the TSCH protocol, allowing parent and channel selection in an intelligent manner. This intelligent allocation aims to optimize the performance of TSCH in the IIoT. The various in-depth simulations carried out show that the integration of QL significantly improves the performance of these networks, including radio reliability, Packet Delivery Ratio (PDR), and especially latency and energy consumption.