Optimized CNNs to Indoor Localization through BLE Sensors Using Improved PSO

Area of Science:

• Computer Science
• Electrical Engineering
• Robotics

Background:

• Indoor navigation is crucial for applications like clinical workflow analysis.
• Bluetooth Low Energy (BLE) beacons and Received Signal Strength Indicator (RSSI) are common for indoor localization.
• Existing RSSI methods require dynamic interference pattern configuration.

Purpose of the Study:

• To explore an alternative method for monitoring a moving user's indoor position using BLE sensors in complex environments.
• To develop a Convolutional Neural Network (CNN) based positioning model for enhanced indoor navigation.

Main Methods:

• A CNN model was developed using a 2D image of signal strength indicators from BLE sensors.
• Neuro-evolutionary approach with enhanced Particle Swarm Optimization (PSO) was used to dynamically optimize CNN layers.
• Dynamic inertia weights were employed in PSO for CNN optimization.

Main Results:

• The proposed optimized CNN-based method achieved high accuracy (97.92%) with a low error rate (2.8%).
• The method demonstrated effective tracking of moving users in a complex building environment.
• Performance was superior to other recent machine learning and deep learning algorithms.

Conclusions:

• The developed CNN-based method offers a highly accurate and efficient solution for indoor positioning.
• This approach simplifies calibration requirements compared to existing RSSI-based methods.
• It provides a robust framework for real-time indoor navigation applications.