A Semantic-Associated Factor Graph Model for LiDAR-Assisted Indoor Multipath Localization

Area of Science:

• Robotics and Autonomous Systems
• Wireless Communication and Signal Processing
• Geomatics Engineering

Background:

• Global Navigation Satellite System (GNSS) is unavailable indoors.
• Wireless signals like 5G and Ultra-Wideband (UWB) are used for indoor positioning.
• Multipath effects from complex indoor structures degrade positioning accuracy.

Purpose of the Study:

• To develop a LiDAR-assisted method for multipath error estimation and high-precision indoor positioning.
• To integrate environmental semantic information for improved multipath correction.
• To enhance positioning robustness in complex indoor environments.

Main Methods:

• A tightly coupled perception-positioning framework using LiDAR and wireless signals.
• A semantic-feature-based neural network for reflective surface detection from LiDAR point clouds.
• A unified factor graph model for joint inference of states and reflector information.

Main Results:

• Accurate extraction of geometric parameters of reflectors using LiDAR.
• Dynamic discrimination and utilization of both line-of-sight (LOS) and non-line-of-sight (NLOS) paths.
• 32.1% improvement in root mean square error (RMSE) compared to traditional methods.

Conclusions:

• The proposed method effectively addresses multipath effects in complex indoor environments.
• LiDAR-assisted semantic information significantly enhances indoor positioning accuracy and robustness.
• Provides a viable solution for high-precision localization in challenging indoor settings.