SSL-SP: A Semi-Supervised-Learning-Based Stream Partitioning Method for Scale Iterated Scheduling in Time-Sensitive Networks
Jingzheng Tu, Qimin Xu, Lei Xu, Cailian Chen · 2021
The demands of reliable and real-time communication in industrial automation systems drive growing attention on Time-Sensitive Networks (TSNs) due to its guarantee of low latency and deterministic transmission. Current works explore iterated scheduling of time-triggered (TT) streams in TSNs mainly by random stream partitioning and graph-based stream partitioning. However, random partitioning obtains limited performance, while the edge weights of graph-based partitioning heavily depend on prior domain knowledge. In this article, we propose a semi-supervised-learning-based stream partitioning (SSL-SP) method for scale iterated scheduling of TT streams in TSNs. SSL-SP automatically discovers stream dependence and clusters streams of similar representation. Besides, we design an evaluation metric for TT stream relevance with no requirements on prior knowledge. Furthermore, we construct a low-rank representation of each stream by sparse encoding. A T2S2dataset on time-triggered stream scheduling is constructed as a comparison benchmark. Simulations demonstrate that SSL-LP achieves the best schedulability rate (86.6%) compared with the state-of-the-art stream partitioning methods.