Enhancing UWB Indoor Positioning Accuracy through Improved Snake Search Algorithm for NLOS/LOS Signal Classification.

Area of Science:

• Robotics and Automation
• Signal Processing
• Machine Learning

Background:

• Non-line-of-sight (NLOS) errors are a primary limitation in ultra-wideband (UWB) indoor positioning systems.
• Accurate distinction between line-of-sight (LOS) and NLOS signals is crucial for improving UWB accuracy.

Purpose of the Study:

• To develop an advanced method for optimizing model parameters to enhance UWB indoor positioning accuracy.
• To improve the classification of LOS and NLOS signals using a refined optimization algorithm.

Main Methods:

• Introduced a novel optimization technique, the "LTSSO-BP" model, by enhancing the Snake Search Algorithm (SSA).
• Incorporated a chaotic map for population initialization and a subtraction-average-based optimizer with dynamic exploration probability into SSA.
• Optimized the initial weights and thresholds of backpropagation (BP) neural networks for signal classification.

Main Results:

• The LTSSO-BP model demonstrated superior stability and accuracy compared to standard BP, PSO-BP, and SO-BP models.
• Achieved high performance metrics: 90% classification accuracy, 91.41% recall, and 90.25% F1 score.
• Effectively distinguished between LOS and NLOS signals, a key factor in UWB positioning.

Conclusions:

• The proposed LTSSO-BP model offers a significant advancement in UWB indoor positioning by improving signal classification accuracy.
• Parameter optimization, rather than network structure modification, is an effective strategy for enhancing UWB positioning performance.
• The refined SSA provides a robust approach for optimizing neural network parameters in signal classification tasks.