Solution for Interference in Hotspot Scenarios Applying Q-Learning on FFR-Based ICIC Techniques

Area of Science:

• Wireless Communication Engineering
• Telecommunications Network Optimization
• Machine Learning Applications in Networks

Background:

• Addresses challenges of varying user concentration and its impact on cellular network performance.
• Explores existing solutions like Multiple-Input Multiple-Output (MIMO) and small cells.
• Introduces inter-cell interference coordination (ICIC) using fractional frequency reuse (FFR) as a key technique.

Purpose of the Study:

• To evaluate the effectiveness of ICIC techniques in mitigating co-channel interference.
• To analyze the impact of high user concentration on system performance.
• To propose and validate a dynamic, machine learning-based solution for optimizing ICIC parameters.

Main Methods:

• Performed exploratory analysis to demonstrate ICIC effectiveness and compare techniques.
• Conducted statistical studies to identify critical system performance parameters.
• Developed and simulated a Q-Learning algorithm within the ns-3 simulator for dynamic ICIC parameter adjustment.

Main Results:

• Demonstrated significant reduction in co-channel interference through ICIC techniques.
• Highlighted the detrimental effects of high user concentration on network performance.
• The proposed Q-Learning algorithm improved average Signal-to-Interference-plus-Noise Ratio (SINR) for all users, with hotspot users experiencing gains of 11.2% to 180%.

Conclusions:

• Machine learning-based dynamic ICIC parameter adjustment is effective in managing time-varying user concentration.
• The Q-Learning approach successfully enhances SINR, particularly in user hotspot areas.
• The proposed solution offers a robust method for maintaining optimal network performance in dynamic cellular environments.