Odometry covariance

[ricber]

Hi Koide,
Can I ask you why you didn't fill the covariance matrix of the nav_msgs::Odometry odometry message? I see that your UKF class exposes a method called getCov(). Could I use this method to fill the covariance of the /odom topic?

Thank you.

[koide3]

Hi @ricber ,
Yes, you can retrieve the covariance using getCov(). But, I expect that the estimated covariance doesn't well represent the true uncertainty of the localization result because the UKF doesn't take scan matching errors into account.

[ricber]

Ok, thank you!

Also, why did you put to zero the linear twist? Because I see you're estimating linear velocities in the Kalman Filter.

hdl_localization/apps/hdl_localization_nodelet.cpp

Lines 257 to 259 in a76067c

	odom.twist.twist.linear.x = 0.0;
	odom.twist.twist.linear.y = 0.0;
	odom.twist.twist.angular.z = 0.0;

[koide3]

Yeah, of course, we can put the estimated velocity in the odom message. The reason I didn't is that the velocity was not necessary in my application, and I was lazy to check the velocity output coordinate...