Multi-Agent Transformer for Queue-Level XR Traffic Scheduling in TSN Networks

2026-08-05

AI Summary

This study addresses the challenge of scheduling heterogeneous, highly dynamic XR traffic in Time-Sensitive Networking (TSN) with shared queues at the edge, where existing approaches lack fine-grained modeling and coordination of inter-queue dependencies. To overcome this limitation, the work proposes a decentralized scheduling method based on multi-agent reinforcement learning, introducing the Multi-Agent Transformer (MAT) to TSN queue-level scheduling for the first time. The approach leverages attention mechanisms to explicitly model dependencies among observations and actions across queues, thereby enabling implicit coordination among heterogeneous XR applications. Experimental results demonstrate that, compared to baseline methods, the proposed solution reduces end-to-end latency by up to 71.42% and decreases task failure rates by as much as 83.2%, while ensuring high reliability across all queues.

Abstract

Time-Sensitive Networking (TSN) and Mobile Edge Computing (MEC) hold strong potential for enabling ultra-reliable low-latency communication for time-sensitive applications, such as eXtended Reality (XR). However, the widespread adoption of XR introduces significant challenges due to co-located services in MEC environments, leading to contention for shared network resources. Moreover, XR traffic types have distinct characteristics and criticality in terms of timing requirements, further increasing the complexity and dynamics of such environments. Although reinforcement learning has shown promise for TSN scheduling optimization in dynamic network scenarios, existing approaches rely on centralized or high-level multi-agent designs and are typically tailored to periodic and predictable industrial traffic, limiting their applicability to XR workloads. As a result, these approaches suffer from (i) limited ability to capture inter-queue dependencies due to coarse-grained control, and (ii) poor adaptability to highly dynamic and heterogeneous XR traffic. To address these gaps, we propose a multi-agent reinforcement learning approach for queue-level XR traffic scheduling. We adopt the multi-agent transformer (MAT) to model inter-queue dependencies via attention over agents' observations and actions, enabling implicit coordination across heterogeneous co-located XR applications. Our simulation results show that the proposed method outperforms baselines, achieving up to 71.42% latency reduction and up to 83.2% reduction in failure rate, while consistently achieving high reliability across all queues.

Problem

• Time-Sensitive Networking
• eXtended Reality
• Multi-Agent Reinforcement Learning
• Queue-Level Scheduling
• Traffic Heterogeneity

Innovation

• Multi-Agent Transformer
• Queue-Level Scheduling
• XR Traffic
• Time-Sensitive Networking
• Reinforcement Learning