Hi all guys!
We tried to use the DWM1001 system (MDEK1001 kit with Decawave PANS) in a supermarket environment. You can find attached the map of the store: each yellow circle is an anchor (each anchor is about 10m away from the adjacent ones and placed at a height of about 5.5m), red,blue and yellow rectangles are the shelves that delimit the lanes (numbered from 1 to 23), blue dots are the test points exploited to evaluate the kit localization performance and black dashed line delimits the evaluation area.
Each shelf has a variable height of up to about 2.5m/3m.
Each test point has been evaluted at 3 different heights along the shelf:
about 0.2m from the floor
about 1.5m from the floor
more than 2m from the floor
In the last setup (i.e. with the tag at more than 2m from the floor) we found a coverage in more than the 95% of the test points (for coverage I mean when the tag can range with at least 3 not-alligned anchors and then log its x,y position) with an average localization error of about 0.8m, that we think is a good result considering the environment.
We found more problems in the other 2 setups (with a lower tag height). Especially in the first one, with the tag height about 0.2m, we can not assure the coverage in about the 50% of the test points, and also the accuracy drops downs, with an average error of about 1.5m/2m (which makes the technology totally unsuitable for the purpose of distinguish each individual lane).
The second setup presents intermediate results between the other two.
So, I have some questions.
First of all, I would like to know if anyone here has already used this technology in a store environment and possibly could give me some feedbacks about his experience.
Then, I would like to know from UWB experts if metal shelves of that height can affect radio frequencies so much that they can cause 2m localization errors or even make the range of the tag with the anchors directly impossible, limiting in this way the coverage at low heights along the lanes.
Clearly, any advice about increasing coverage and accuracy in our setup and for our purpose is welcome, especially about increase coverage at low heights along the shelves.
I hope I have explained my problems succesfully, I did my best.
For any further clarification, do not have problem to let me know, I will try to explain better my setup and my encountered problems.
Thank you for your time and consideration!
Very interesting experiment. Based on many posts here, my own experience in my own office and the Notes in Decawave papers I realized awhile ago that although supermarket or any store for this purpose look like a great place for this technology it might be very tricky because of metal shelves interference. Avoiding it probably requires installing anchors on the ceiling which is way more complex and expensive - takes more time and measuring equipment…
Now, I am not sure why would you measure the tags at 0.2m… I thought that this kind of system gives the management and marketing better idea of people traffic in the isles, not what particular products they pick, which would be a wishful thinking…
Avoiding it probably requires installing anchors on the ceiling which is way more complex and expensive - takes more time and measuring equipment…
We projected to increase anchors height (we would like to place them at 7m instead of 5.5m, the ceiling is more than 10m and it is really prohibitive to place anchors there), to see if maybe we will have better performance also at low heights along the shelves.
Now, I am not sure why would you measure the tags at 0.2m… I thought that this kind of system gives the management and marketing better idea of people traffic in the isles, not what particular products they pick, which would be a wishful thinking…
The purpose is tracking goods position for stocking goal, so we need to know in which lane they are, also the ones that are stocked in the lowest part of each shelf…
I know that could be very challenging, but I feel that the key point is to decrease as much as possible the shadow cone of the shelves respect to UWB anchors.
I am waiting for forum users feedback about this idea to create a brainstorming as complete as possible.
We have a test on a site with many metal shelves. Many readings come in with distance too large, but none with distance too small. Since the tags are in staff pockets, we have concluded this is due to the signal reflected from a shelf behind being much stronger than the direct path attenuated by the person.
For what I know about UWB, you are experiencing classic NLOS effect. There could be an obstacle (maybe staff body or metal shelves) in between tag and anchor, that leads to an increasing time-of-flight measuarement. NLOS effect is always evident with a measured range that is greater than the real one.
