Re: 3D-map is so bad

Posted by matlabbe on
URL: http://official-rtab-map-forum.206.s1.nabble.com/3D-map-is-so-bad-tp8867p8880.html

Hi,

In ENU, z acceleration should be positive +9.81 when standing still on a flat surface (so that measured acceleration +9.81 + gravity -9.81 = 0). The TF frame of the imu is maybe inverted. Normally with IMU, robot_localization yaw estimation should be quite accurate, and with /gx5/imu/data it was pretty bad (I don't know why as I don't know how its quaternion is computed). You can show the velodyne_points with some history in rviz with Odom fixed frame set in Global options. If odom is accurate, the velodyne_points should accumulate like my second screenshot of my previous post, but was accumulating like the first screenshot.

The madgwick filter takes the angular velocity and linear acceleration data from the input IMU msg and computes the quaternion. In this case it recomputes the quaternion of /gx5/imu/data into output /imu/data fed to ekf in my example. The quaternion from /imu/data makes more sense than the one in /gx5/imu/data, so ekf gives good odometry.

cheers,
Mathieu