RPLIDAR distortion/calibration

[Thomas Marguet]

Joe,

Something I noticed running some tests indoor is that my RPLIDAR seems to have some sort of concave distortion before the addition of the window. See below a slice of my living room... definitely the ceiling and floor are not that curved. This is processed without windo corner and window distortion parameters in the *.imu file 

 

Is that something you observed or is it my rplidar getting more doggy? If yes, maybe the window calibration should take this into consideration as the window calibration adjusts the scan through the window to a potential no window distorted scan. 

Cheers

Thomas

[Thomas Marguet]

I think I might have the explanation... I forgot to remove the window distortion parameters from the imu config... I'll run more tests but I guess it should be ok. Sorry for the inconvenience.

Thomas

[Thomas Marguet]

Hi Joe,

Back on this matter after further tests.

The distortions observed were not related to the window distortion parameters. I'm exploring Szymon's lidar alternative and therefore I dismantled the lidar head. I might have damage the lidar... or not.

When I plotted the "no window" calibration scan of a section of a room (i.e. rectanglar shape), I observed a clear distortion : instead of being rectangle, the ceiling and floor have a concave shape, and the distance measured are significantly higher than the reality. 

I investigated the possibility of a misalignement of the laser (i.e. not exactly 90° of the rotation axis) but I couldn't get rid of the curves and the distances observed are too far off to be caused by a misalignement. I reassembled the rplidar as it was originally and I still observed a similar results... so the issue doesn't seem to be caused by the alternative lidar assembly or a laser misalignment. 

The issue has to be with the lidar: the lidar is damaged or the lidar is not as accurate as the specs says. 

I looked into the few scans I acquired indoor before dismantling the lidar. I think the distance distortion was already present but I don't have a wide enough scan to be 100% sure as the distortion only become obvious beyond 2m. I then plotted the distance error vs actual distance (alternative assembly/no window) and the graph suggests there's a linear relation. Using the equation given by Excel to correct the distance measured gives pretty good results.

Do you observe the such distance distortion with your regular lidar when scanning without window ? 

Thomas

[caveatron-discuss]

Thomas,

I'm sorry that you are having problems. Of the 3 RPLIDARs I have worked with, none had distortion that bad. Yes, they do have a tiny amount but you have to look very closely to see it - nothing like what you are seeing. As an example, here is a screenshot of raw data straight out of the LIDAR with no window and no corrections applied. (The two weird spots on the ceiling are light fixtures and you can even see the baseboards by the floor.)

Joe

[Thomas Marguet]

Hi Joe,

I did see your data in the window calibration spreadsheet and so the lights... Your calibration room is about 3,5m wide, which I think isn’t enough to assess the possible distance distortion. Up to 2m range, the distortion isn’t much noticeable. I did my test in a 9 m long corridor, and then the distortion became obvious. I’m not saying my lidar is ok but it worth trying on a much longer range.

I have ordered 2 new rplidar so I should be able to see if the issue remains. I won’t be able to test it before last week of August. 

If there it appears the issue remains, it should be possible to correct the error either way in acquisition or post-processing phases.

Thomas

--
You received this message because you are subscribed to the Google Groups "caveatron-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to caveatron-disc...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/caveatron-discuss/0186accf-d370-4322-8f4e-8ec1268e8015n%40googlegroups.com.

[Thomas Marguet]

Hi Joe

I received the 2 new RPLIDAR A1M8 devices and tested them. I see nothing similar to what I observed with the first RPLIDAR. Still I can't tell if the issue was there orignialy or not.

Unfortunately, the 2 devices received have there own issue... one (red) as the reference axis misaligned by 2° which could be corrected in the lidar rotation parameter but I don't like it... and the second one (white) has oscillating patterns which I suppose could be cause by the rotating part being unbalanced... :( . Checking with the supplier what can be done.

Thomas