Deep Reinforcement Learning-Based Adaptive Scheduling for Wireless Time-Sensitive Networking.

Area of Science:

• Computer Science
• Electrical Engineering
• Networking

Background:

• Time-Sensitive Networking (TSN) is crucial for real-time communication.
• Adapting TSN to wireless IEEE 802.11 networks presents challenges like channel contention and dynamic conditions.
• Existing TSN schedulers struggle with wireless latency and variability.

Purpose of the Study:

• To propose a wireless TSN model for exclusive channel access in IEEE 802.11 networks.
• To develop a novel time-sensitive traffic scheduler, WISE, utilizing deep reinforcement learning.
• To address latency issues and ensure reliability in wireless TSN.

Main Methods:

• Developed a deep reinforcement learning (DRL) framework to model time-sensitive traffic patterns.
• Designed and implemented the Wireless Intelligent Scheduler (WISE) algorithm.
• Conducted experiments to evaluate WISE's performance in diverse wireless scenarios.

Main Results:

• The proposed WISE algorithm achieved up to 99.9% reliability across various wireless scenarios.
• Processing delays were successfully limited within specified time requirements.
• The scalability of TSN streams was effectively guaranteed by the proposed mechanisms.

Conclusions:

• The wireless TSN model and WISE scheduler offer a robust solution for low-latency wireless communication.
• Deep reinforcement learning is effective in optimizing TSN performance in dynamic wireless environments.
• The approach ensures high reliability and meets strict timing constraints for critical applications.