An Intelligent Handover Management System for 5G Networks Using A Hybrid Machine Learning Approach

Abstract

With the development of fifth-generation (5G) wireless networks, it is now possible to have ultra-reliable and low-latency communication, which has been used in autonomous vehicles, smart cities, telemedicine, and the Internet of Things (IoT). One of the issues in the 5G networks is how to deal with continuous transitions, especially in high mobility dense urban areas with small-cell deployment. Widely used threshold-based handover systems which depend on measures like Reference Signal Received Power (RSRP) and Signal-to-Interference-plus-Noise Ratio (SINR) tend to be unsatisfactory leading to more latency, handover failures and ping-pong effects. This paper proposes a smart handover management system of 5G networks based on a hybrid machine learning framework that integrates the Long Short-Term Memory (LSTM) networks with the Reinforcement Learning (RL) learning frameworks. The LSTM model anticipates the states of the network in the future and user mobility pattern, offering insights into the variability in signals, whereas the RL agent makes the best handover decisions involving the predicted and real-time network state. The hybrid model is developed in Python and trained in the Google Colab platform, and system-level analysis in MATLAB/Simulink. The key performance metrics such as the handover latency, handover failure rate, the ping-pong rate, the throughput and the loss of packets are measured and compared with the traditional threshold-based techniques. Simulation results indicate that the proposed system reduces handover latency by approximately 37%, decreases handover failures by over 60%, lowers ping-pong events by ~63%, improves throughput by up to 27%, and reduces packet loss by 64-68%. The presented improvements indicate that proactive, predictive handover management in which ML is applied can be highly beneficial in terms of improving the network reliability and Quality of Service. The paper sheds light on the promise of hybrid predictive-decision systems to optimise handovers in dynamic 5G systems, which provides a viable solution to the high-speed, low-latency applications in the future.