Hi all,

We are currently working on the integration of a navigation algorithm to autonomously fly a quadrotor indoor. We are currently using the hectorslam node offered by ROS to get a estimate of the quadrotor position. We also use the laserheight_estimation to measure the altitude and obtain a full 3D pose.

We would like to use the robotposeekf to merge inertial data coming from our IMU (3dm-gx3-25) with our SLAM measurement to make our navigation system more robust.

I have finished the implementation but the filter update itself very slowly (around 0.1hz) when the other sensors are feed asynchronously at 40hz (eventually, the IMU could be sent at 200hz). So I think I am doing something wrong but what?

I have some idea of where are my mistakes, so I got also those questions :

1) Which input topic should I use? Right now, I have renamed odom to odom2D (which is not used), I publish my SLAM pose into the vo topic and the IMU into the imu_data. I have set the correct settings in my config files.

2) Should I use the covariance return by the hector_slam (the values seems very high regarding data that is relatively really precise).

3) The hectorslam (and laserheight_estimation) doesn't return any linear velocities. What should I do? If I understand correctly, I should but high value in the covariance matrix?

4) Does hector_localization is a better option ?