Optimized network slice allocation in 5G using Q-learning-based reinforcement learning

The proliferation of 5G networks demands intelligent resource allocation techniques to meet diverse Quality of Service (QoS) requirements across various application domains. This paper presents a Q-learning-based reinforcement learning approach to dynamically allocate network slices—eMBB, URLLC, and mMTC—based on real-time parameters such as packet loss, delay, and Guaranteed Bit Rate (GBR). Unlike traditional rule-based methods, our framework leverages experience-based learning to adapt slice allocation in dynamic environments. The environment simulates multiple network states, applies a reward function aligned with QoS goals, and derives optimal policies using tabular Q-learning. Experimental results demonstrate over 90% accuracy in slice selection, validating the model’s efficiency and suitability for real-time deployment in 5G/6G RAN orchestration scenarios.