Dynamic Time-Sensitive Resource Allocation via Reinforcement Learning in 5G Slicing with Edge-Aware QoS Guarantees

**Abstract:** This paper proposes a novel approach to dynamic resource allocation within 5G network slicing architectures, leveraging Reinforcement Learning (RL) to optimize time-sensitive network (TSN) performance. Existing resource allocation strategies often rely on static configurations or reactive adjustments, failing to proactively adapt to fluctuating network conditions and stringent TSN QoS requirements. Our method, termed Edge-Aware Dynamic Resource Allocation with Prioritized TSN (EADR-PTSN), integrates edge computing capabilities and prioritizes TSN flows based on criticality, achieving significantly improved latency, jitter, and throughput compared to traditional scheduling algorithms. The system's self-learning capacity allows it to optimize resource allocation strategies in real-time, maximizing network utilization while guaranteeing strict TSN QoS constraints, paving the way for enhanced real-time industrial applications.