Publication: HYBRID ARTIFICIAL INTELLIGENCE SCHEME FOR VERTICAL HANDOVER IN HETEROGENEOUS NETWORKS
0
0
Issued Date
2024-12-12
Resource Type
Scopus ID
2-s2.0-105005958952
Journal Title
ACM International Conference Proceeding Series
Start Page
47
End Page
53
Rights Holder(s)
SCOPUS
Bibliographic Citation
ACM International Conference Proceeding Series (2024) , 47-53
Suggested Citation
Phatcharasathianwong S., Kunarak S. HYBRID ARTIFICIAL INTELLIGENCE SCHEME FOR VERTICAL HANDOVER IN HETEROGENEOUS NETWORKS. ACM International Conference Proceeding Series (2024) , 47-53. 53. doi:10.1145/3694875.3694884 Retrieved from: https://hdl.handle.net/20.500.14740/21022
Author(s)
Author's Affiliation
Corresponding Author(s)
Other Contributor(s)
Abstract
This comprehensive study investigates the optimization of handover processes in heterogeneous networks (HetNets) utilizing a hybrid approach combining Long Short-Term Memory (LSTM) and Double-Q Learning (DQL). The research is pivotal for modern telecommunication networks that demand high-speed, efficient, and seamless data transfer, especially for users in motion. The primary objective is to minimize unnecessary handovers and enhance overall network efficiency in transitioning connections among LTE, 5G, and Wi-Fi 6 technologies. The methodology integrates DQL with LSTM to create a robust decision-making framework that dynamically adapts to user movement and network conditions. This hybrid model significantly reduces the probability of service disruptions and data transmission inefficiencies typically encountered in dense network environments. The approach leverages the LSTM’s capability to predict optimal handover points based on sequential data analysis and the DQL’s effectiveness in refining these predictions to prevent ping-pong effects and radio link failures. In a simulated scenario within a 100x100 meter area, encompassing a 20-story building with 1,000 randomly distributed users (20% outdoors and 80% indoors), the network’s performance under various conditions, including users moving at speeds of 0-10 km/hr., was meticulously analyzed. The simulation employed advanced machine learning techniques to manage the network’s service radii effectively—15 meters for Wi-Fi6, 35 meters for 5G, and 1,500 meters for LTE—ensuring optimal coverage and connectivity. Results from the study indicated a total of 603 handovers, with a notable success rate of 94%. This outcome highlights the efficacy of integrating LSTM with DQL in reducing unnecessary handovers, enhancing signal strength detection, and improving system throughput. The study demonstrated that this innovative approach could effectively mitigate 36 instances of ping-pong effects and completely manage to avoid critical radio link failures in three instances. This research underscores the transformative potential of applying sophisticated computational models to solve complex problems in network management. By addressing key challenges such as latency, packet loss, and the efficiency of the handover process, the study not only enhances user experience by ensuring continuous service during mobility but also supports bandwidth-intensive applications like live streaming and online gaming. These advancements herald significant improvements in network reliability and performance, paving the way for future innovations in telecommunications.