Optimization of Device-Free Localization with Springback Dual Models: A Synthetic and Analytical Framework

Area of Science:

• Signal Processing
• Wireless Communication
• Localization Technologies

Background:

• Traditional device-free localization (DFL) methods using received signal strength (RSS) struggle with accuracy and efficiency in complex environments due to multipath effects and noise.
• Existing convex sparsity regularization methods are computationally convenient but fail to capture signal sparsity effectively.
• Non-convex methods offer better sparsity approximation but suffer from high computational complexity and local optima issues.

Purpose of the Study:

• To propose novel synthetic models for device-free localization (DFL) that overcome the limitations of traditional methods.
• To introduce a weakly convex penalty function (Springback) that balances sparsity promotion and signal amplitude preservation.
• To develop an efficient Springback-transform model for large-scale data processing in DFL.

Main Methods:

• A novel synthetic model utilizing a weakly convex penalty function, Springback, combining ℓ1 compression and ℓ2 rebound terms.
• A Springback-transform model based on analytical transform learning for direct sparse feature extraction.
• Solving both models using a difference of convex algorithm (DCA) to enhance computational efficiency.

Main Results:

• The proposed Springback models significantly improve positioning accuracy and computational efficiency in DFL.
• Experimental results show high accuracy and low positioning error across various complex environments.
• The models outperform existing state-of-the-art DFL methods in terms of performance and computation time.

Conclusions:

• The developed Springback models offer a robust solution for device-free localization in challenging environments.
• The novel approach effectively addresses the trade-offs between accuracy, efficiency, and computational complexity in DFL.
• These findings present a practical advancement with significant potential for real-world DFL applications.