| Summary: | The supervised descent method (SDM) avoids computing inverse of the Hessian matrix and is a potential tool to rapidly solve the nonlinear least squares problem of magnetic dipole localization. However, the magnetic measurements are often noisy, which will cause an error during the update of SDM. To address this issue, we proposed a singular value truncated SDM (TSDM) to seek the descent directions that have the greatest differences in magnetic intensities. The results of the simulations and the experiment show that: 1) TSDM is more robust than SDM and obtains localization errors comparable or lower than Levenberg-Marquardt (LM) and 2) TSDM is faster than LM when the number of sensors ≤ 25 and the signal-to-noise ratio (SNR) ≤ 30 dB. Thus, the proposed TSDM may help to build a robust and fast magnetic localization system.
|
|---|
