DynaTAS: Dynamic Time-Aware Shaper Communications in a 5G-TSN Converged Network for Remote Healthcare

With the advent of Industry 5.0, applications such as Smart healthcare are offering unprecedented services such as remote patient monitoring, robotic surgery, etc. These services require reliable, low-latency communications for patient safety and precision. The IEEE Time Sensitive Networking (TSN) standards address determinism in communications through Time Aware Shaper (TAS) scheduling of time-triggered traffic. But this is primarily available in wired interfaces. A converged 5G-TSN network is capable of offering wireless support for TAS in smart healthcare services. However, communication determinism in terms of bounded latency and jitter becomes challenging in the wireless domain due to radio channel access and unpredictable traffic flows. In this paper, we propose DynaTAS, a Reinforcement Learning (RL) algorithm to compute efficient time-aware schedules that are adapted based on network dynamics. DynaTAS uses a closed control loop in which the RL agent continually reconfigures the TAS schedules in response to instantaneous variations in the queue states. We validate through simulation that the RL-based approach successfully optimizes TAS configurations, achieving the bounded latency and stable jitter, which is critical in meeting stringent smart healthcare requirements. DynaTAS was extensively evaluated on OMNeT++ simulations and was able to reduce the latency by up to 95.7% and the jitter by up to 98.7% compared to a static baseline.