Time of Flight (ToF) VL53L5CX - 8x8 pixel sensor

@inq EDIT Oops that was CDN, US is $10. I think the price went up, it was on DigiKey but now it's $14.79.

Posted by: @zander

OR remind me why CV and a normal camera can't do a better job?

You may need to remind me what CV is, but video doesn't give me depth information without triangulation AND the big one is recognizing which two pixels represent the same item out in space... so you can do that triangulation.  The computations are just too difficult to do accurately without serious CPU.  I do not feel that using reference points like targets and things on the ceiling are viable either.  I don't want to have to do a bunch of stuff in a new room JUST so the robot can work.  I want to simply put it in a room (and eventually house, building) and it map the whole thing... 1200 sq-ft -  100,000 sq-ft. Fire and Forget.  

Inherently, I don't see the 8x8 resolution as a hindrance.  More pixels like a rotating scanner just means more has to be analyzed or more have to be dumped.  The 8x8 just came along and it give me a systematic grid pattern and I still believe it will do as I want. I may be optimistically delusional 🤣 but I'm happy in my insanity.  

Posted by: @zander

@inq I am fairly sure the max range is only 400CM/4M, what spec is telling you otherwise?

I'm not sure what this is referring to.  Yes this VL53L5CX sensor is rated to 4m.  If you are referring to the earlier private message, I am considering getting one of the single point lidar units that has lens on it, that is good out to 12 meters.  

VBR,

Inq

Posted by: @zander

@inq EDIT Oops that was CDN, US is $10. I think the price went up, it was on DigiKey but now it's $14.79.

Like I said... I don't think that one is on a break-out board.  That company sells to everyone... hobbyist as well as manufacturers and I can't work with something this tiny.  I see nothing on their site showing a break-out board version.

@inq Ok, I was going by a picture I saw here. Just to be clear, this one is too small? That's the $10 unit.

That's the beast... but look at the dimensions.  It's just a bit bigger than a rice grain.

@Inq I am not sure if it helps you, but Apple has an app that works with the iPhone 12 and up which has built in Lidar. You could maybe use the phone to create a map and then load it into the robot. At least with Apple using it the price should now come down with volume increasing.

Posted by: @zander

@Inq I am not sure if it helps you, but Apple has an app that works with the iPhone 12 and up which has built in Lidar. You could maybe use the phone to create a map and then load it into the robot. At least with Apple using it the price should now come down with volume increasing.

 

I don't drink the Apple Kool-Aid!

@inq I am sure Android will have it as well.

BTW, on AliExpress a Lidar unit is $10.70CDN, can't get US price plus shipping. 

@inq I didn't realize it was so small.

Posted by: @zander

AliExpress a Lidar unit is $10.70CDN

You and I are on opposite sides of the Amazon/AliExpress coin.  But, maybe its time for me to give them ANOTHER chance.  If you would be so kind???... it took me ten minutes to get to your post here.  I tried to start up Ali Express five minutes ago... still nobody home (Verizon's Fault).  A search on Ali Express might take me an hour.  Next time, could you provide a link.  If you feel that's too much like marketing on the forum, please send to me private.  I would greatly appreciate it.  Thanks.  

VBR,

Inq

@inq 

You may need to remind me what CV is, but video doesn't give me depth information without triangulation AND the big one is recognizing which two pixels represent the same item out in space... so you can do that triangulation.

Open Source Computer Vision Library.

The computations are just too difficult to do accurately without serious CPU.

Visual processing is computationally intensive.

I do not feel that using reference points like targets and things on the ceiling are viable either. I don't want to have to do a bunch of stuff in a new room JUST so the robot can work. I want to simply put it in a room (and eventually house, building) and it map the whole thing... 1200 sq-ft - 100,000 sq-ft. Fire and Forget.

Fair enough. Some vacuum robots use vision making use of natural features to map and navigate a house.

Inherently, I don't see the 8x8 resolution as a hindrance. More pixels like a rotating scanner just means more has to be analyzed or more have to be dumped.

But you will be moving your sensor around collecting a lot more data than a simple rotating scanner which would go beyond the 4m limit. Navigating with local features is not as good as using distant features visible from any position. You seem to want a 3d cloud map? Where is it to be stored? Lots of data there and no mention of how it is going to be processed? The speed your little bot moves about will require a fast response time and that is not going to happen with the sensor you have except for things like obstacle avoidance.

@inq

I am out in town, will get you a link when I get back to my laptop 

Posted by: @robotbuilder

But you will be moving your sensor around collecting a lot more data than a simple rotating scanner which would go beyond the 4m limit.

Very true... and why I'm looking at a single beam lidar unit.  I think you are right from the accuracy and tighter beam.  There is less chance of having to figure out if it is seeing a near and far image at the same time.  Bill's video you posted sometime back didn't showcase lidar unit... just another ToF sensor as an alternative to ultrasonic.  Which, I completely agree with what we've said here... it is only a few steps better than ultrasonic and only in the right conditions.

Anyway, I'm looking at something like this unit, but would greatly appreciate anyone with ANY opinion...

"https://www.amazon.com/gp/product/B09B6Y2GBJ"

https://www.amazon.com/gp/product/B09B6Y2GBJ

VBR,

Inq

Posted by: @robotbuilder

 

Inherently, I don't see the 8x8 resolution as a hindrance. More pixels like a rotating scanner just means more has to be analyzed or more have to be dumped.

But you will be moving your sensor around collecting a lot more data than a simple rotating scanner which would go beyond the 4m limit. Navigating with local features is not as good as using distant features visible from any position. You seem to want a 3d cloud map? Where is it to be stored? Lots of data there and no mention of how it is going to be processed? The speed your little bot moves about will require a fast response time and that is not going to happen with the sensor you have except for things like obstacle avoidance.

I don't know about that.  A rotating sensor does what... a couple hundred samples per rotation and rotates how many RPM... I have to inspect or throw away 90% of that data.  If I want anything above/below the horizontal plane of the device, I have to angle it and half of the data is totally useless.  With this sensor, I can get an organized 8x8 grid.  I don't have to run it at the 15 Hz and in fact, I think if I start doing the higher integration scans, I have to trigger them manually and thus, I can turn the head to where I want and take the snapshot.  

My thoughts based on my testing observations AND on this thread (Thanks guys, you've been great AND very constructive) is...  that I can do a quick circle of the room in sort of a "survey" mode that gets the basic dimensions, but as you pointed out at the extremes it get's more randomness.  Maybe this randomness is because of the sensor or in areas where things are busy... table legs and such.  I'm hoping I can identify those areas and have the robot go back to those areas for closer-up, thorough scans.  AGAIN... this is totally at the conceptual, brain-storming level of thought.

As far as the data... you all convinced me it was not cheating to do things off-bot.  So, the design of Inqling Jr was totally based on a nimble, data collector...  No real brains.  My belief if I can get the statistics under my belt... is to take the raw data cloud as it's coming in to the real computer and FIND geometric planes... so, hundreds (or thousands) of points become a simple plane equation.  The robot in the future can't go through a plane.  As you brought up earlier... game "hit detection" works by this method.  Point being... what might be a Gig of data at the computer should easily get condensed down to even the 3MB available on an ESP8266.  Heck... a plain rectangular room of 4 wall planes (without furniture, etc) would take a whopping 96 Bytes Total!  I feel even a measly Arduino Uno could handle driving around with this kind of map system!

Just my aspiration.  😉 😊 😎 

VBR,

Inq

@inq 

Bill's video you posted sometime back didn't showcase lidar unit...

Have you viewed this video?

https://dronebotworkshop.com/getting-started-with-lidar/