Which quadcopter to choose for inventory? And how do you take inventory?

Question

Barcodes

Interesting experience in choosing and advice from those who have already encountered.
The task is this:
1. Conduct an inventory in the warehouse.
2. Automatic flight.
3.Which readers to use (barcode) RFID tags.
4. What quadrocopter is needed for such purposes.

Answer 1

Um... If you have a barcode that is not a meter by a meter and in a clearly defined place, then a useless exercise (positioning a quadric in a room is not an easy task) - I have a bad experience with this.
From the reader - only the camera, because. most readers need a short distance, there are more professional ones, but they also have a decent weight along with the price.
Therefore, instead of a pocket quadric, it turns into a full-fledged Phantom.
It is better, of course, that there would be Qr codes - it is easier to look for them in the picture.
We once developed an attempt to automate with a friend. Mechanics on the principle of the Delta printer, only on 2 cables moved through the rows. (I don’t know the subtleties of mechanics, I did only the software part).
Well, in the end, this was abandoned, because. one person with a portable makes faster, tk. if the box on the shelf is turned on the wrong side, then you still need to turn it over.
It all depends on the task that needs to be solved, in my opinion.
Our tasks were as follows:
1. Reduce the number of "losses" (mainly due to the fact that the movers put them on the wrong shelf)
2. Reduce the number of thefts from boxes (namely, when the goods were taken out of one common box).
In general, both problems were solved at once by installing weight sensors under each place with a mandatory scan of the box when loading into place.

Answer 2

I'm afraid the companies capable of such experiments in the world can be counted on the fingers, I don't think that you will find ready-made software. If there is no one in the warehouse
during the inventory (an important point) and you are ready to carry out manual scanning in case of noticeable changes (copter with lidar is controlled manually), then you can try to pick up hardware with the following features - obstacle detector and indoor navigation. As a payload, the copter should have a lidar (with proper red-eye, you can save money and add an Intel 3d camera, there are models of 1-10 meters and millimeter accuracy) or a high-quality camera with a memory card.
Don’t even dream about full automation, everything will look like this - you manually scan the room (you can manually or control a copter with lidar), then, based on this scan, you collect an accurate map and, based on it, manually build a flight plan. The copter must be able to fly according to the plan and automatically return to the base if it detects an obstacle and cannot fly around it (this will create a problem in the software, but solvable).
After the flight, you can take the collected video, it can already be scanned on the machine, combining the time and the flight plan, you can identify barcodes on the image.
ps separately, all of the above can be used by any copter with the required payload capacity, with external automated control (navigation is only external, based on a video surveillance system, (any student can detect a light bulb on a copter in a dark room on video), but I'm afraid the cost of integration can cost a pretty penny (I strongly recommend doing it on your own) precisely because there are no ready-made offers on the market.
pps a room scanner based on an intel camera there was an article on habré, it even seems that the software is bundled with sdk as an example, but the scan will need to be corrected at key points, since the error accumulates as the scan progresses, in general, a lot of manual work.
You will have to scan the room more than once! even an extra box can create problems, so ideally you need to create software so that a special person walks through it with a camera and a laptop in the evening after the warehouse closes, marking key points, then uploads the collected scan to the editor, corrects it and, based on it, rules the flight plan. How to do all this, no idea, but at least you can find / write a plug-in for a 3D editor, the same sketchup.
With good quality of video surveillance cameras, it is possible to partially automate the decision on the need to rescan the premises, it is possible to combine old and new images of the premises, obtaining a map of changes from the delta (it is necessary to pre-mark the images on the floor and shelves), i.e. the operator can rescan not the entire room, but only the changed parts.
By the way, fuck the copter! fuck the flight plan! this is all unnecessary work, initially the task was set incorrectly!
Every day, a specially trained person walks through the warehouse, pushing a cart with a laptop and 5-10 cameras that look to the sides (the cart - because the warehouse can be high) and record all the boxes passing by. Passing each next rack, the operator presses a magic button that looks for the nearest mark on the wall, linking the video to the code on it.
Then all this video is loaded onto a powerful machine, analyzed, labels or even images on the boxes are scanned, if necessary, you can recognize the text (you have no idea what can be done from a high-quality photo), .. what to do with this data, think for yourself. Coordinates can be determined relative to key marks and inertial navigation, or count the revolutions of the wheels of the trolley (your choice or combine).
Pluses - no need to scan the premises, no need for quadcopters, etc. the warehouse premises are somehow inspected at the end of the working day, let the person with the cart do it. Another advantage is the ability to use several cameras instead of one (it is also a minus - since the amount of data for processing may be too large).

Answer 3

If you have RFID tags, then why quadcopter, flight and cameras?
The RFID scanner can read passive RFID tags up to 2 meters (HF) or up to 10 meters (UHF and microwave). That is, just smuggle the scanner between the racks and that's it.
If you need to fully automate, then on the racks there are rails (for example, from a curtain rod) + Raspberry Pi + motor + scanner.

Answer 4

Hello, I'm working on a similar solution. I would like to know more about the requirements for the system, maybe one day we can solve your problem.