MAIL Positioning System Provides a Localization Boost for Indoor Navigation Using Magnetic Readings

A team of computer scientists at Sun Yat-sen University, the Guilin University of Electronic Technology, and the University of Connecticut have released details of a new indoor navigation technique based on geomagnetic sequences: MAIL, the Multi-scale Attention-guided Indoor Localization system.

"Knowing accurate indoor locations of pedestrians has great social and commercial values, such as pedestrian heat-mapping and targeted advertising," the team explains in the paper's abstract. "Location estimation with sequential inputs (e.g., geomagnetic sequences) has received much attention lately, mainly because they enhance the localization accuracy with temporal correlations."

"Nevertheless, it is challenging to realize accurate localization with geomagnetic sequences due to environmental factors, such as non-uniform ferromagnetic disturbances. To address this, we propose MAIL, a multi-scale attention-guided indoor localization network, which turns these challenges into favorable advantages."

The MAIL system is effective three novel approaches wrapped up into one. Firstly, the team designed a scale-based system for feature extraction which accounts for variations in readings at different scales. Secondly, the attention-generation part of the system identifies attention values for different scales adaptively. Finally, these scale-based attention values are fused into multi-scale features for location estimation.

The result, the researchers found, is a dramatic improvement in positional accuracy: Across three trial sites — a lab area, a food court, and a parking lot, through which a user walked with an off-the-shelf smartphone collecting magnetic readings — the team found that MAIL reduced localization errors by over 36 percent compared to current indoor positioning schemes.

The paper has been published under open-access terms as part of the ACM International Joint Conference on Pervasive and Ubiquitous Computing 2020 (UbiComp '20).