Noise covariance from sensors are 0....?

asked 2021-10-04 23:55:05 -0500

kidpaul
38 ●11 ●17 ●20

updated 2021-10-04 23:58:58 -0500

Hello, I'm very new in Kalman filter and I'm trying to use EKF packages (i.e., robot_pose_ekf or robot_localization) to fuse odometry and IMU in the HSR from Toyota Research Institute. When I first tried to hook the robot_pose_ekf package with my mobile robot, I encountered this following error:

[ERROR] [1519539033.564366334]: Covariance specified for measurement on topic wheelodom is zero
[ERROR] [1519539036.363531683]: filter time older than odom message buffer
....

So I checked the published data from the odometry (nav_msgs/Odometry) and IMU (sensor_msgs/Imu), and the covariance matrices in those are indeed all 0.

I am confused since I don't know whether it is common to have 0 covariances from those published data and we need to characterize and estimate those sensor noise covariances by collecting bunch of data with trials and errors (I don't have any sensor specification though)....?

edit retag flag offensive close merge delete

add a comment

3

answered 2021-10-05 01:06:36 -0500

gvdhoorn
86574 ●283 ●1432 ●1054 http://cor.tudelft.nl/

updated 2021-10-05 01:18:54 -0500

There is no way for generic components to be aware of the specific characteristics of your hardware / sensors, so yes, if you have a specific platform, and the ROS installation was not provided by the OEM / mfg, it's likely covariances (and similar properties) will be blank/zero/empty (unless whoever wrote the sensor/hw drivers has done work to identify them, and has configured those components to also publish them).

Seeing as the HSR is a platform, I'd be surprised though if some other user / group has not figured those out already. You might want to ask around on a HSR-focused forum or support channel. Those values may not be exactly right for your specific robot, but they would provide a good starting point.

edit flag offensive delete link

Comments

Yeah, that's what I was expecting.... I should probably ask this to the company since they officially provide ROS package for this robot.

kidpaul ( 2021-10-05 15:42:52 -0500 )edit

add a comment

0

answered 2022-12-19 14:23:56 -0500

kidpaul
38 ●11 ●17 ●20

updated 2022-12-19 14:26:23 -0500

Short Updates:

According to Toyota Research Institute (supplier of the official ROS package of the HSR), they do not have this information with them unfortunately.

edit flag offensive delete link

add a comment

Noise covariance from sensors are 0....?

2 Answers

Comments

Question Tools

Stats

Related questions

Noise covariance from sensors are 0....? edit

2 Answers

Comments

Question Tools

Stats

Related questions

Noise covariance from sensors are 0....?