Why there are negative values in the diagonal of covariance matrix when I use laser_scan_matcher?

asked 2016-05-25 02:33:20 -0500

DaveD gravatar image

updated 2016-05-26 00:39:54 -0500

commands list

$ roslaunch laser_scan_matcher demo.launch publishe_covariance:=true

$ rostopic echo /pose_with_covariance

and the outputs are:

---
pose: 
  position: 
    x: 0.549596190337
    y: -0.178588658868
    z: 0.0
  orientation: 
    x: 0.0
    y: 0.0
    z: -0.149123030231
    w: 0.988818649629
covariance: [7.090325225361744e-07, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -1.683608329440047e-07, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1e-09, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1e-09, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1e-09, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -8.930799963827599e-08]
---
pose: 
  position: 
    x: 0.549030877813
    y: -0.178521699171
    z: 0.0
  orientation: 
    x: 0.0
    y: 0.0
    z: -0.148700666534
    w: 0.988882253745
covariance: [6.980619750991466e-07, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -1.2904810885632077e-07, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1e-09, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1e-09, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1e-09, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -7.140501191520443e-08]
---
pose: 
  position: 
    x: 0.548898420892
    y: -0.178495689681
    z: 0.0
  orientation: 
    x: 0.0
    y: 0.0
    z: -0.149505424992
    w: 0.988760905324
covariance: [7.045827487110208e-07, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -1.8801387061818686e-07, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1e-09, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1e-09, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1e-09, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -1.0193466813256203e-07]

so my question is why there are negative values in the diagonal of covariance matrix?

edit retag flag offensive close merge delete

Comments

1

They are also extremely low. Your standard deviation would be below 1mm which seems very overconfident.

Felix Endres gravatar image Felix Endres  ( 2016-05-25 04:50:04 -0500 )edit

This is a demo from laser_scan_matcher package, and the data used are provided with the package, not my data. However the covariance matries are weird. I want to know how to handle this kind of situation.

DaveD gravatar image DaveD  ( 2016-05-26 00:21:58 -0500 )edit

Hi, i have the same issue, even when i go through a corridor, the covariance still very low...do you have an idea about this situation ? Thanks

Lotfi_Z gravatar image Lotfi_Z  ( 2019-05-16 16:23:16 -0500 )edit