Revision history [back]

Even though "Fixed Frame" says "odom" in your picture, rviz seems to display the point cloud in the coordinate frame of the sensor. I.e. your sensor is placed at the displayed coordinate axes.

You need to set the "Fixed Frame" to a frame that is (in reality) placed on the floor and make the transformation from there to the sensor known via tf. Often these transformations are available from the robot model and published on /tf, so you need only to select the correct frames. If you attached the sensor yourself, you probably need to tell rviz the transformation from, e.g., the gripper to the sensor's optical center (which would be an extrinsic calibration issue).

Then fixed frame in RViz should be, e.g. "base_link", "odom", "world" or "map". The "frame_id" set in the header of the pointcloud then needs to be the frame of the sensor.