[large covariance]hectorslam(hokuyoUTM30lx)+rtabmap(kinectV2)

[seoho0527]

I'm using
1) Hector slam with Hokuyo UTM30lx lidar
2) Rtabmap with kinect v2

I'm passing odometry from hector slam to rtabmap.

During that process hector slam covariance is too large for rtabmap to get.

1) How can I tune covariance from hector slam?
2) How can I make rtabmap filter out large covariance?
3) Hector slam has /poseupdate & /slam_out_pose(doesn't include covariance)
How can I only pass /slam_out_pose to rtabmap and make this process work.

[matlabbe]

Hi,

For hector slam, I cannot really help about tuning their covariance. However, it is possible to fix odom covariance manually in rtabmap to bypassing input ones. To do so, you have to use odometry from TF by setting odom_frame_id parameter of rtabmap node. Then to fix covariance, set odom_tf_angular_variance and odom_tf_linear_variance parameters (this will fill diagonal of the 6x6 covariance matrix):

<node pkg="rtabmap_ros" type="rtabmap" value="rtabmap">
   ...
   <param name="odom_frame_id" value="odom"/>
   <param name="odom_tf_linear_variance" value="0.0001"/> <!-- ~1 cm error -->
   <param name="odom_tf_angular_variance="0.0005"/>       <!-- ~1.3 deg error -->
</node>

cheers,
Mathieu