Intelligent Traffic Scheduling in Time-Sensitive Networks Based on Deep Reinforcement Learning

Time-sensitive Networking (TSN) is a deterministic networking technology designed for time-critical applications. Some traffic in TSN must comply with certain hard real-time conditions, so traffic scheduling is the core of TSN to ensure low latency and determinism. Traditional scheduling approaches primarily include solver-based methods and heuristic algorithms. However, solver-based methods suffer from high computational complexity, while heuristic methods require manual design tailored to specific application scenarios and rely heavily on domain knowledge. Therefore, there is an urgent need for an efficient TSN scheduling method. In this study, we propose a scheduling method based on deep reinforcement learning (DRL), RTAIS — Routing and Time Agents for Intelligent Scheduling, which splits the traffic scheduling problem into two sub-problems of routing and time planning, and intelligent agents are used to achieve efficient scheduling in TSN. Simulation results show that RTAIS schedules nearly twice more flows than heuristic algorithm on average, and increases the scheduled flows by 22.92% on average compared to the existing deep reinforcement learning method.