Learning-Based Traffic Classification for Mixed-Critical Flows in Time-Sensitive Networking

TL;DR

Traffic-Type Assignment ($TTA$) in Time-Sensitive Networking ($TSN$) is NP-hard when guaranteeing hard real-time ($HRT$) deadlines while maximizing soft real-time ($SRT$) QoS. The authors formulate the problem as an optimization and propose a Proximal Policy Optimization ($PPO$)-based DRL model, $TTASelector$, integrated with Gate Control List (GCL) generation via ASAP and network calculus (NetCal) for Worst-Case Delay ($WCD$) estimation. The work introduces an end-to-end framework that jointly handles topology, flow criticality, utility functions, NetCal analysis, and scheduling generation, and demonstrates that $TTASelector$ achieves 100% $HRT$ schedulability and up to ~95% $SRT$ QoS on realistic topologies with runtimes of a few seconds, vastly outperforming a Tabu-search metaheuristic that requires hours. This approach enables scalable, automated TSN configuration in mixed-critical networks, reducing manual intervention and providing a foundation for further learning-based TSN optimization.

Abstract

Time-Sensitive Networking (TSN) supports multiple traffic types with diverse timing requirements, such as hard real-time (HRT), soft real-time (SRT), and Best Effort (BE) within a single network. To provide varying Quality of Service (QoS) for these traffic types, TSN incorporates different scheduling and shaping mechanisms. However, assigning traffic types to the proper scheduler or shaper, known as Traffic-Type Assignment (TTA), is a known NP-hard problem. Relying solely on domain expertise to make these design decisions can be inefficient, especially in complex network scenarios. In this paper, we present a proof-of-concept highlighting the advantages of a learning-based approach to the TTA problem. We formulate an optimization model for TTA in TSN and develop a Proximal Policy Optimization (PPO) based Deep Reinforcement Learning (DRL) model, called ``TTASelector'', to assign traffic types to TSN flows efficiently. Using synthetic and realistic test cases, our evaluation shows that TTASelector assigns a higher number of traffic types to HRT and SRT flows compared to the state-of-the-art Tabu Search-based metaheuristic method.

Figures (5)

• Figure 1: Contribution and overall workflow of the paper.
• Figure 2: Utility Function (D represents the soft deadline of the SRT flows, and BD represents the buffer deadline).
• Figure 3: Environment workflow for flow assignment and reward calculation.
• Figure 4:
• Figure 5: