Advanced Handover Optimization (AHO) using deep reinforcement learning in 5G Networks
Abstract Handover (HO) management in 5G networks is an essential and sensitive mechanism as the deployment of 5G networks is undergoing rapid changes. We propose Adaptive Handover Optimization (AHO) model that uses Deep Reinforcement Learning (DRL) to dynamically adapt those key Handover Control Par...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-07-01
|
| Series: | Journal of King Saud University: Computer and Information Sciences |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s44443-025-00124-0 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Handover (HO) management in 5G networks is an essential and sensitive mechanism as the deployment of 5G networks is undergoing rapid changes. We propose Adaptive Handover Optimization (AHO) model that uses Deep Reinforcement Learning (DRL) to dynamically adapt those key Handover Control Parameters (HCPs) to increase the handover completion rate and the request service rate via finetuning the Handover Margin (HOM), Time to Trigger (TTT). In this model, out of three paramount parameters, namely Handover Probability (HOP), Outage Probability (OP) and Ping Pong Handover Probability (PPHP), the main issue is to minimize signal degradation and service interruption via simultaneous optimization of these three parameters. These metrics are performance indicators when the HO process is being evaluated. Thus, a Deep Deterministic Policy Gradient (DDPG) based Actor Critic framework with four Deep Neural Networks (DNNs) is proposed to train four DNNs to intelligently adjust HOM values based on the real time Radio Signal Strength (RSRP) and Signal to Interference plus Noise Ratio (SINR) conditions. The random mobility of User Equipment (UE) in the distributed base station coverage zone is modelled by the system environment, which subsequently derives HO decisions. Simulation proves that the model achieves a significant reduction of unnecessary handovers and outage probabilities, leading to improve the stability of the network. The proposed model contributes a scalable, learning based optimization strategy applicable to systems of future generation wireless communication. |
|---|---|
| ISSN: | 1319-1578 2213-1248 |