A Reinforcement Learning Handover Parameter Adaptation Method Based on LSTM-Aided Digital Twin for UDN.

Area of Science:

• Telecommunications Engineering
• Artificial Intelligence in Networks
• Wireless Communication Systems

Background:

• Optimizing network performance in ultra-dense networks (UDNs) is challenging due to complex handover dynamics.
• Effective handover parameter adaptation is crucial for maintaining seamless connectivity and user experience.
• Existing methods often struggle with dynamic changes in wireless signal fading conditions.

Purpose of the Study:

• To propose a novel deep Q-learning (DQN) method for dynamically selecting handover parameters in UDNs.
• To enhance the efficiency and performance of the DQN algorithm using Long Short Term Memory (LSTM) for digital twin modeling.
• To optimize the effective handover ratio by adapting parameters to real-time wireless signal fading.

Main Methods:

• A deep Q-learning (DQN) algorithm is employed to dynamically adjust handover parameters.
• Long Short Term Memory (LSTM) networks are utilized to construct a digital twin for assisting the DQN.
• The proposed method dynamically selects parameters based on wireless signal fading conditions for backward compatibility.

Main Results:

• The enhanced DQN-LSTM method demonstrates a faster convergence speed compared to the standard DQN.
• An average increase of 2.7% in the effective handover ratio was achieved.
• The proposed method exhibits superior performance across different wireless signal fading intervals.

Conclusions:

• The integration of LSTM-based digital twins significantly enhances the DQN algorithm's efficiency in UDNs.
• Dynamic adaptation of handover parameters based on signal fading conditions improves network performance.
• The proposed approach offers a promising solution for optimizing handover in challenging wireless environments.