Using Machine Learning to Speed Up the Design Space Exploration of Ethernet TSN networks

In this work, we ask if Machine Learning (ML) can provide a viable alternative to conventional schedulability analysis to determine whether a real-time Ethernet network meets a set of timing constraints. Otherwise said, can an algorithm learn what makes it difficult for a system to be feasible and predict whether a configuration will be feasible without executing a schedulability analysis? In this study, we apply standard supervised and unsupervised ML techniques and compare them, in terms of their accuracy and running times, with precise and approximate schedulability analyses in Network-Calculus. We show that ML techniques are efficient at predicting the feasibility of realistic TSN networks and offer new trade-offs between accuracy and computation time especially interesting for design-space exploration algorithms.