Throughput Optimization of IEEE 802.15.4e TSCH-Based Scheduling: A Deep Neural Network (DNN) Scheme

Abstract

This paper describes a simple and reliable deep learning-based deep neural network (DNN) model that can conduct time-slotted channel hopping (TSCH) based scheduling in accordance with IEEE 802.15.4e guidelines. In a centralized fashion, the TSCH network develops as a maximum weighted bipartite matching strategy for links to cell assignment of a slot frame. The cell assignment problem is solved using a well-known Hungarian assignment algorithm, which considers network throughput as a bipartite-edge-weight. We use the Hungarian scheduling technique to create training data and train the DNN accordingly. The results of the simulations show that the proposed DNN-based scheduling scheme outperforms Hungarian algorithm-based methods while using fewer computational resources.