Indoor Pedestrian Navigation Based on Conditional Random Field Algorithm.

Area of Science:

• Robotics and Autonomous Systems
• Geomatics Engineering
• Signal Processing

Background:

• Foot-mounted micro-electromechanical systems (MEMS) inertial sensors are crucial for pedestrian navigation and indoor localization.
• Existing systems struggle with accumulated heading errors, leading to significant drift over time.
• Previous zero-velocity detection algorithms correct positional errors but not heading drift.

Purpose of the Study:

• To develop a novel map-matching technique to mitigate heading errors in pedestrian navigation systems.
• To enhance the long-term accuracy of indoor localization using inertial navigation systems (INS).
• To reduce the computational complexity of the localization algorithm.

Main Methods:

• A map-matching technique based on conditional random fields (CRFs) was proposed.
• Observations were defined as positions from the INS with consistent lengths between consecutive points.
• A simplified CRF model with four states and a single heading-based feature function was employed.
• A sliding window with a feedback structure was integrated to boost accuracy.

Main Results:

• The proposed algorithm demonstrated efficient improvement in long-term accuracy.
• Experiments conducted over 450 m in two different sites validated the algorithm's effectiveness.
• The simplified approach reduced algorithmic complexity while maintaining high performance.

Conclusions:

• The CRF-based map-matching technique effectively addresses heading errors in MEMS-based pedestrian navigation.
• This method offers a significant improvement in the long-term accuracy of indoor localization systems.
• The reduced complexity makes the algorithm practical for real-world applications.