@mapjournals.bsky.social#MAV
Local-Tukey outlier removal with an improved iterative ensemble Kalman filter cuts four-anchor indoor UWB localization RMSE from 0.1588 to 0.1451 m under multipath and NLOS.
Details: https://www.maxapress.com/article/doi/10.48130/mav-0026-0011
#UWB #KalmanFilter #IndoorLocalization
@daniel.colquitt.xyzTap-water nitrate T is modelled separately as a slowly varying Gaussian random walk. That is an assumption, but a more plausible one than treating the supply as perfectly constant.
[4/10]
@daniel.colquitt.xyzThis is the same mathematics as the “nearly constant velocity” model used in target tracking: N takes the role of position and r of velocity.
[3/10]
@daniel.colquitt.xyzMy hidden state is
x = (N, r, T)
with r the net accumulation rate and T the nitrate concentration of the tap water.
[2/10]
@daniel.colquitt.xyzI have, perhaps inevitably, built a Gaussian assumed-density filter (GADF) to estimate nitrate in my #Aquarium .
I initially thought of it as a #KalmanFilter , but parts of the physical and measurement model are non-Gaussian, so GADF is the more accurate description.
@daniel.colquitt.xyz.ap.brid.gyI have, perhaps inevitably, built a Gaussian assumed-density filter (GADF) to estimate nitrate in my #Aquarium.
I initially thought of it as a #KalmanFilter, but parts of the physical and measurement model are non-Gaussian, so GADF is the more accurate […]
