ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
0

LIDAR produces "phantom walls"

asked 2017-09-08 11:55:36 -0500

Pi Robot gravatar image

Hello,

We have a SICK LMS100 LIDAR attached to the front of our mobile robot and we have been using it for several years. We suddenly started noticing "phantom walls" appearing in the laser scan when viewed in RViz. You can see a video of the phenomenon here. The small yellow dots in the video are produced by a rotating LIDAR on the head of the robot and these dots line up correctly with the surrounding walls and other obstacles. The larger colored dots are the LMS LIDAR points shown in intensity view. Near the beginning of the video you can see how the LMS points line up with the yellow dots along a wall at the far right. As the robot backs up, the LMS readings "pull off" the real wall but continue to produce a phantom wall that moves with the robot.

We have observed the same phenomenon whenever there is a symmetrical gap that the robot is facing and the phantom wall is always positioned at roughly the same distance from the robot.

Can anyone explain what is going on here? Is the LMS faulty at the hardware level? Or is this some kind of reflection problem that we can filter?

Thanks! patrick

edit retag flag offensive close merge delete

Comments

Is the lidar getting tilted and is pointing at the floor in those instances? Try putting something on the floor with some height but that should always lie beneath the scan plane but will be noticeable if the robot or sensor angles down.

lucasw gravatar image lucasw  ( 2017-09-08 12:37:54 -0500 )edit

Maybe the lidar lens is dirty or the power input to the lidar is drooping? P.S. I recognized your name on the Youtube channel, I read ROS by Example a while back, excellent book!

Airuno2L gravatar image Airuno2L  ( 2017-09-08 13:34:44 -0500 )edit

I agree. I use the book and recommend it to others.

jayess gravatar image jayess  ( 2017-09-08 17:50:28 -0500 )edit

@lucasw - Thanks for the suggestion but the scanner is definitely not tilting down and data further away than the "phantom" wall shows up OK.

Pi Robot gravatar image Pi Robot  ( 2017-09-08 17:56:56 -0500 )edit

@Airuno2L - Wouldn't a dirty lens cause a permanent reading rather than the odd intermittent reading we get? Also, we can reproduce the phantom wall effect at certain locations time after time so I don't think it is a power fluctuation which presumably would occur at other locations as well.

Pi Robot gravatar image Pi Robot  ( 2017-09-08 18:00:24 -0500 )edit

P.S. Thanks for the book recommendation. :)

Pi Robot gravatar image Pi Robot  ( 2017-09-08 18:00:54 -0500 )edit

Can you share a bag of what is in that video on botbags or marvhub?

lucasw gravatar image lucasw  ( 2017-09-08 18:46:19 -0500 )edit

I'll post a bag file on Monday assuming we haven't figured out the issue by then.

Pi Robot gravatar image Pi Robot  ( 2017-09-09 06:52:56 -0500 )edit

2 Answers

Sort by ยป oldest newest most voted
0

answered 2017-09-11 22:51:35 -0500

Pi Robot gravatar image

And the answer is...Tilt as suggested by @lucasw. It turns out the bracket holding the LIDAR had been very slightly bent downward. Once we bent it back level, the problem went away. The weird thing is (and it's what confused us the most), is that the downward tilt did not cause a constant reading from the floor a certain distance out from the robot. Instead, it only produced the "phantom wall" under very specific conditions, namely, when the view ahead included a narrowing at a certain distance, like two pillar supports of a wall jutting into the floor space. Move the robot a little forward or backward and the fake wall would go away.

Anyway, problem solved.

edit flag offensive delete link more
1

answered 2017-09-10 23:50:00 -0500

I had the exact same effect when placing the SICK LMS100 too close to the floor in my robot (the infrared light does not travel in a perfect beam, it becomes wider with distance. So placing it too close to the floor will cause the sensor to be really sensitive to tilt and see the floor. You would be surprised how uneven floors really are). In the video I see the effect happening more often when the robot is moving backwards. That suggests me that it is actually tilting a bit (in my experiments less than 1 degree is enough to produce the effect). The problem was solved by placing the sensor higher in my robot's structure.

edit flag offensive delete link more

Question Tools

2 followers

Stats

Asked: 2017-09-08 11:55:36 -0500

Seen: 589 times

Last updated: Sep 11 '17