A Systematic Flexible-Window based Scheduling Framework for Time-Sensitive Networking

Time-Sensitive Networking (TSN) is increasingly applied in automotive and industrial internet fields due to its low latency and deterministic communication. Gate control list (GCL) is foundational for deploying TSN. Currently, most scheduling research focuses on frame-to-window based scheduling. This scheduling approach typically generates a specific window for each frame, leading to a proliferation of GCL in large networks, which increases the complexity of implementing TSN. To simplify deployment and enhance scheduling reliability, this paper introduces a systematic flexible-window based scheduling framework. Utilizing a gapless GCL design approach, it optimizes flow's worst-case end-to-end (e2e) delays through window length design, with delays obtained through network calculus analysis. A generic solving framework based on metaheuristic algorithms is established to address this optimization problem. The scheduling framework also features a load-balanced turn prohibition routing strategy to balance link loads and avoid cyclic dependencies, alongside a K-means priority clustering method based on routing overlap to reduce the number of priorities. Simulation validation in a high-level autonomous driving vehicle's in-vehicle network shows that the proposed method can decrease GCL numbers by nearly 90% against frame-to-window scheduling. In common industrial internet scenario, it significantly reduces worst-case e2e delays and enhances scheduling success rates compared to the analogous scheduling method. Large-scale complex network scenario further demonstrates its scalability. © 2025 IEEE.