Due the the inte­gra­tiv and recur­sive nature of the cus­tom­ary ZUPT-aided INS, the covari­ances will grow through­out each ZUPT-less period and col­laps as soon as ZUPTs are applied. This will give rise to large cor­rec­tions giv­ing ‘dis­con­ti­nu­ites’ in the esti­mated tra­jec­to­ries. How­ever, for many appli­ca­tions, such as motion analy­sis and visu­al­iza­tion, some lag can be tol­er­ated and there­fore smooth­ing can be applied to remove the dis­counti­nu­ities. We have ini­tial­ized some work on smooth­ing for ZUPT-aided INS. Hope­fully we will have some arti­cles on the sub­ject soon, but in the mean­while we will make the cur­rent sta­tus of the code avail­able here.

Smooth­ing Matlab

Just to try it and see the results, you can down­load the file, unzip it and run main.m in Matlab.

Acknowl­edge­ment: The work on smooth­ing has pri­mar­ily been car­ried out by our master-thesis stu­dent David Simón Colomar.

Comments are closed.