Dynamic Spectrum Access Algorithms Based on Survival Analysis

Area of Science:

• Wireless communication
• Signal processing
• Data science

Background:

• Dynamic spectrum access (DSA) is crucial for efficient radio spectrum utilization.
• Traditional DSA methods often struggle with accurately predicting spectrum availability.
• Survival analysis offers a novel approach to model time-to-event data, applicable to spectrum availability.

Purpose of the Study:

• To design and implement two novel dynamic spectrum access algorithms.
• To leverage survival analysis for predicting remaining idle spectrum time.
• To evaluate algorithm performance in real-world scenarios and assess cross-dataset applicability.

Main Methods:

• Developed two algorithms based on non-parametric survival analysis (cumulative hazard function).
• Predicted remaining idle time for secondary transmissions with a specified success probability.
• Validated algorithms using Long Term Evolution (LTE) band data to model primary user activity.
• Tested algorithms across different training and testing dataset combinations.

Main Results:

• Algorithms demonstrated effectiveness in real-world scenarios, even at fine time scales.
• Performance remained robust when algorithms were trained on one dataset and applied to another, provided similar cumulative hazard functions.
• Achieved high white space utilization.
• Measured probability of interference remained at or below the preset threshold.

Conclusions:

• The proposed survival analysis-based algorithms are effective for dynamic spectrum access.
• The algorithms offer reliable spectrum availability prediction and efficient spectrum utilization.
• Cross-dataset training demonstrates the algorithms' adaptability and robustness in dynamic radio environments.