Autonomous Integration of TSN-unaware Applications with QoS Requirements in TSN Networks

TL;DR

The paper addresses enabling QoS guarantees for TSN-unaware applications in TSN networks by introducing SCIP, an autonomous proxy that signals QoS requirements on behalf of legacy streams. A Deep Recurrent Neural Network (DRNN) is used to detect periodic traffic, followed by a traffic-description extraction stage that yields the TSN descriptor $D=(w,m,f_{max})$ and a QoS-class determination via traffic classification (e.g., Rhebo Industrial Protector). The contributions include the SCIP architecture with four processing stages, a DRNN-based periodicity detector, a heuristic for automatic traffic description, and an evaluation showing improved QoS for a TSN-unaware VoIP stream under overload, along with compatibility analysis with TSN shaping/policing. This work enables legacy and TSN-unaware streams to benefit from TSN QoS without modifying end-hosts, and demonstrates practical viability through OMNeT++ simulations and detailed method validations. The approach offers a path toward automatic, scalable QoS integration in industrial networks while preserving existing TSN mechanisms.

Abstract

Modern industrial networks transport both best-effort and real-time traffic. Time-Sensitive Networking (TSN) was introduced by the IEEE TSN Task Group as an enhancement to Ethernet to provide high quality of service (QoS) for real-time traffic. In a TSN network, applications signal their QoS requirements to the network before transmitting data. The network then allocates resources to meet these requirements. However, TSN-unaware applications can neither perform this registration process nor profit from TSN's QoS benefits. The contributions of this paper are twofold. First, we introduce a novel network architecture in which an additional device autonomously signals the QoS requirements of TSN-unaware applications to the network. Second, we propose a processing method to detect real-time streams in a network and extract the necessary information for the TSN stream signaling. It leverages a Deep Recurrent Neural Network (DRNN) to detect periodic traffic, extracts an accurate traffic description, and uses traffic classification to determine the source application. As a result, our proposal allows TSN-unaware applications to benefit from TSNs QoS guarantees. Our evaluations underline the effectiveness of the proposed architecture and processing method.

Figures (9)

• Figure 1: The fully centralized configuration model is composed of end stations, bridges, one or more CUCs, and a single CNC. IEEE P802.1Qdj 8021qdj defines the communication interface between CUC and CNC.
• Figure 2: Architecture of a simple feedforward NN. The input data travels through the network in one direction, without any loops. Each Neuron calculates the weighted sum over all inputs and outputs the corresponding value of the activation function $\theta$.
• Figure 3: Extended configuration model in accordance with the fully centralized model. The SCIP examines network traffic and invokes stream reservation on behalf of applications. Further, it configures the first bridge on the path to apply 802.1CB stream identification methods.
• Figure 4: Processing stages within the SCIP for a TSN-unaware stream. The input for the processing pipeline are recorded network packets and the output are the parameters required for a TSN stream announcement.
• Figure 5: The dataset consists of 4 different kinds of streams: aperiodic, near-periodic, periodic, and periodic patterns. Aperiodic and near-periodic streams should be classified as aperiodic. Periodic streams and periodic patterns should be classified as periodic.