An Indoor Navigation Algorithm Using Multi-Dimensional Euclidean Distance and an Adaptive Particle Filter.

Area of Science:

• Robotics
• Computer Science
• Geomatics Engineering

Background:

• Inertial Navigation Systems (INS) offer high short-term accuracy but suffer from cumulative drift.
• WiFi fingerprinting avoids cumulative error but is prone to signal mismatching.
• Integrating INS and WiFi fingerprinting is a common approach to leverage the strengths of both.

Purpose of the Study:

• To develop an improved indoor localization algorithm by combining INS and WiFi fingerprinting.
• To address the limitations of cumulative error in INS and mismatching in WiFi fingerprinting.
• To enhance the accuracy and reliability of pedestrian dead reckoning systems.

Main Methods:

• Utilized a particle filter with dead reckoning as the state transfer equation and the difference between INS and WiFi fingerprinting as the observation equation.
• Incorporated floor map information to eliminate particles that cross walls.
• Proposed an adaptive particle filter that adjusts particle weights based on trust levels and historical data.
• Introduced a multidimensional Euclidean distance algorithm to mitigate WiFi fingerprint mismatching.

Main Results:

• The adaptive particle filter effectively assigns higher weights to more reliable particles.
• The multidimensional Euclidean distance algorithm significantly reduces WiFi fingerprint mismatching.
• Experimental results demonstrate substantial improvements in positioning accuracy compared to traditional methods.

Conclusions:

• The proposed integrated system effectively overcomes the individual limitations of INS and WiFi fingerprinting.
• The adaptive particle filter and multidimensional Euclidean distance algorithm contribute to high-accuracy indoor localization.
• This approach offers a robust solution for accurate and reliable navigation in complex indoor environments.