A Novel Outdoor Positioning Technique Using LTE Network Fingerprints.

Area of Science:

• Wireless communication
• Machine learning
• Geospatial positioning

Background:

• Wireless fingerprint positioning is crucial for Internet of Things (IoT) applications.
• Existing methods face challenges in outdoor environments, particularly with signal instability.

Purpose of the Study:

• To propose a novel deep learning-based fingerprint positioning approach leveraging Long-Term Evolution (LTE) signals for outdoor environments.
• To enhance the robustness and accuracy of deep neural networks (DNNs) for positioning.

Main Methods:

• LTE signal measurements converted into location grayscale images to form a fingerprint database.
• Implementation of cross-entropy loss, dynamic learning rate adjustment, and data enhancement techniques for DNN robustness.
• A hierarchical training method using deep residual networks (Resnet) and transfer learning (Feed-Forward Neural Network - FFNN) for coarse and fine localization.

Main Results:

• The proposed DNN automatically learns location features from LTE signals.
• The hierarchical training approach with Resnet and FFNN-based transfer learning yields a more accurate fine localizer.
• Satisfactory outdoor positioning accuracy was achieved, demonstrating the effectiveness of the deep learning model.

Conclusions:

• The developed deep learning model effectively addresses outdoor positioning challenges using LTE signals.
• The image-based fingerprinting and hierarchical training strategy significantly improve positioning accuracy.
• This approach offers a promising solution for reliable outdoor positioning in IoT ecosystems.