A Bluetooth Indoor Positioning System Based on Deep Learning with RSSI and AoA.

Area of Science:

• Wireless communication
• Indoor positioning systems
• Machine learning

Background:

• Traditional Received Signal Strength Indicator (RSSI) and Angle of Arrival (AoA) methods suffer from multipath effects, signal attenuation, and noise in complex indoor environments, limiting positioning accuracy.
• Existing indoor positioning technologies require further enhancement to overcome environmental challenges and achieve reliable performance.

Purpose of the Study:

• To develop a robust deep learning-based Bluetooth indoor positioning system.
• To mitigate the impact of environmental interference on positioning accuracy.
• To improve upon the performance of traditional and emerging indoor positioning methods.

Main Methods:

• Utilized a Kalman filter (KF) to reduce angular error in Angle of Arrival (AoA) measurements.
• Employed median filter (MF) and moving average filter (MAF) to mitigate fluctuations in Received Signal Strength Indicator (RSSI)-based distance measurements.
• Proposed a deep learning network architecture combining a Convolutional Neural Network (CNN) with a Multi-Head Attention (MHA) model, trained using the backpropagation (BP) algorithm.

Main Results:

• The proposed system achieved an average positioning error of 0.29 meters.
• Demonstrated significant accuracy improvements compared to traditional RSSI and AoA positioning methods.
• Exhibited superior positioning accuracy relative to various emerging indoor positioning techniques.

Conclusions:

• The developed deep learning-based Bluetooth indoor positioning system effectively overcomes the limitations of traditional methods.
• The integration of advanced filtering techniques and a CNN-MHA model enhances positioning accuracy in complex indoor environments.
• This approach offers a promising solution for high-precision indoor localization.