New magnetic navigation algorithm cuts indoor positioning error by 46%

Researchers in China developed a magnetic-inertial navigation method that uses a new 3D magnetic field model to improve indoor positioning where GNSS, Wi-Fi and Bluetooth are unavailable. Tested on a public dataset, the system reduced horizontal error to below 1.27 meters and outperformed a prior method by 46% on average.

Why it matters: - Indoor navigation still breaks down in places where satellite signals cannot reach, including tunnels, parking garages and steel-heavy buildings. - The new method could make navigation more reliable for firefighters, warehouse robots, autonomous vehicles and mine operations. - The approach also reduces dependence on costly infrastructure such as dense Wi-Fi or Bluetooth networks.

What happened: - Researchers from the Aerospace Information Research Institute, Chinese Academy of Sciences published a magnetic-inertial odometry method in Satellite Navigation on June 5, 2026. - The method is called FSS-EMD-MIO, short for Fibonacci sphere-sampled equivalent magnetic dipole model-based magnetic-inertial odometry. - The system combines an array of 30 small magnetometers with an inertial measurement unit to track movement without external signals. - The paper is identified by DOI 10.1186/s43020-026-00201-3.

The details: - The model represents an indoor magnetic field as a set of virtual equivalent magnetic dipoles instead of using smooth polynomial curves. - The team found 16 dipoles to be the optimal number through parameter analysis. - Fibonacci sphere sampling places the dipoles evenly in 3D space and avoids directional bias. - Each dipole’s magnetic moment is solved in real time with least squares fitting. - The researchers derived the spatial gradient of the model to connect changing magnetic readings with displacement, velocity and attitude. - An Adaptive Error State Kalman Filter fuses inertial and magnetic observations to manage nonlinearity and location-dependent noise. - On a public dataset, the method achieved horizontal positioning RMSE below 1.27 meters. - The result beat the previous state of the art, MAINS, by 46% on average.

Between the lines: - The advance targets a key weakness in map-free magnetic navigation: polynomial models can capture broad field trends but miss local anomalies from pipes, boxes and other metal structures. - A more physically interpretable model could make magnetic navigation easier to deploy in real buildings because it better reflects how magnetic disturbances behave in 3D space. - The authors’ emphasis on loop closure and scan matching suggests the current system still needs help with long-term drift before it can support full-scale SLAM.

What's next: - The team plans to add loop-closure detection to correct long-term drift. - Future work will use scan matching based on overlapping magnetic field regions. - The longer-term goal is a complete magnetic SLAM system for multi-floor buildings. - The researchers say that would help close the gap between indoor and outdoor navigation reliability.

The bottom line: - A new magnetic-inertial model turned indoor magnetic noise into a usable navigation signal, cutting positioning error nearly in half without relying on GNSS or fixed infrastructure.