Joint Time-Frequency Resource Scheduling Over CQF-Based TSN-5G System
Zongrong Cheng, Dong Yang, Ruibin Guo, Weiting Zhang
Abstract
Deterministic communication across the integrated time-sensitive networking (TSN) and 5G system (5GS) relies on the queue model and network resource scheduling. In this paper, we investigate the joint time-frequency resource scheduling problem in CQF-based TSN-5GS aiming to improve the number of schedulable flows and system load balance. To support the integration of CQF-based TSN and 5GS, we propose a time-aware packet forwarding strategy for 5GS. By managing the flow transmission time intervals and frequency resources, 5GS can seamlessly communicate with nearby CQF enabled TSN bridges. Given the complexity of network resource scheduling across the TSN-5GS, it is difficult for traditional algorithms to obtain the optimal scheduling results. Hence, we design a DRL-assisted time-frequency resource scheduling algorithm, namely DTF, to jointly allocate the time interval and frequency resources in the unit of flow. DTF embeds with a two-dimensional matrix of link capacity utilization to efficiently learn the relationship between different time intervals and capacity utilization. Experimental results show that our proposed DTF algorithm can increase the schedulable flow number by 54.2% and 3.5x than DeepCQF and Naive, respectively. DTF can also achieve a good trade-off between runtime and schedulability.