Calibrating ST ToF sensors?

[Mark Womack]

Has anyone else out there found the need to calibrate the ST family of ToF sensors? Specifically I am playing with the VL6180, and I have found that the values from the readings can be off fro each other quite a bit, enough to make me worry about the accuracy affecting my code's decision making.

Right now I am only playing in single shot mode, not continuous. So, cross sensor feedback shouldn't be a problem. I found some instructions in the ST VL6180 manual that talks about manually calibrating, but it is a pain I'd like to avoid if I can.  But maybe I can't.

-Mark

[Michael Wimble]

I’m using 8 VL53L0X sensors and they definitely don’t agree with each other on distances. I haven’t worried about calibration so far as I’m using them merely as proximity sensors and a coarse reading is fine. My intent is to provide rings of protection on the bot. The LIDAR and stereo cameras and the SLAM algorithm and path planner should keep things more than, say, a foot from my robot. But, things get missed, especially close in. So my 4 sonars should warn if things get within, say, 8 inches of the robot and the TOF sensors should warn if things get within, say, 4 inches of the robot. I’ll use all 12 proximity sensors identically, but the angles that they cover are different. The sonars a measuring broad angles from the center of the 4 sides of the robot. The TOFS are measuring a narrower beam from the corners of the robot, looking directly front/back and left/right from the corners. The TOFs are at a different height than the sonars.

I find that the TOF sensors are pretty erratic in open space and are only useful when objects get within about a meter. I’ll come back and look at how I can deal with that again, maybe in a couple of months.

On my next design, I may move the TOF and sonar sensors back towards the center of the robot so that they can detect objects pretty close to the robot, since all the sensors have significant minimum distances.

[Steve " 'dillo" Okay]

I'm using VL53L0X sensors as "Cliff" sensors for Tenacity. They're now mounted inside a cover with small slots cut out for the sensor elements, with the actual mount-point recessed back a bit inside to give them some shading.

I get pretty accurate readings from them with this configuration, even outdoors. Maybe shading them a bit might help ?

'dillo

[Mark Womack]

So, I played around with calibrating the sensors (using the ST manual as a guide), and as I played with the sensor it became more accurate. I think that over time the sensor seems to self-calibrate or something. My calibration didn't accomplish very much. Shading them might be a good idea, but I am mostly playing indoors for the tablebot challenges. I'm going to start writing the "find and push an object off the table" behavior and see how the sensors do. I really only need to find which sensor has the lowest value, and then move in that direction. I can move the bot back and forth to make sure of the reading.

I have 10 VL6180s on Beatrice, 2 on each side, 3 front, 3 back. It might be a little overkill, but when I designed it I was thinking that I would also have it do the micromouse challenge, the maze with actual walls instead of a line. It would use the sensors to guide it down the "hallways" and around corners. Beatrice is an 'all around' challenge bot. :-) But a distance sensor being 2 centimeters off could have consequences. Learn by doing, I'll see how it goes.

-Mark

--
You received this message because you are subscribed to the Google Groups "HomeBrew Robotics Club" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hbrobotics+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hbrobotics/e04e721b-eaae-4a6e-a77d-1c7668f851afn%40googlegroups.com.