Proximity Detector

New to Decaware, looking for some advice.

I want to use the tags (preferably without any anchors) to build a proximity detector. Such that, each tag vibrates or sends a GPIO output when it gets within a certain range of any other tag.

Is this something that could be easily done on the Decawave platform? Is there another product that would be better suited for this application? Is this something that has been done before, or that I can buy off the shelf?

I appreciate any advice or feedback given.

It can certainly be done.

You can send out a broadcast and all other units in range respond with their IDs. You can then measure your range to each of those IDs in turn. By playing with the transmit power you could prevent tags well outside of the required range from responding and so reduce the number of ranges you would need to measure.

Exact details of the protocols used would determine the maximum ranges, number of supported tags and update rates, there would be a trade off between these three.

I’m not aware of any end products that would perform this off the shelf so some custom firmware / software would be required.

Depending on the exact requirements BLE bluetooth devices would be another option, lower battery use but far less accurate on the range.

Hi Andy, thanks for enlighting such a good solution and it sounds pretty straightforward.
I m new to Decawave too, would you tell me which example should I look at first for implementing this peer to peer detection?
I prefer to use EVK1000 as a start point. Any suggestions?

B.R.

For that to work, all tags have to be in receive mode all the time they aren’t transmitting. That’s the only way, using only UWB, they can find out about neighbors they need to range to.

That means battery life is very poor. It takes about two AA alkaline cells to run 15 hours for a DW1000. An 18650 LiIon cell could run 24 hours.

In other words, if you have a big battery you are willing to recharge every day, you can solve this problem with UWB.

Dozens of companies claim they have this solved, the no infrastructure UWB social distancing application. I’ve yet to see one that actually works. This is one of those use cases which has a huge gap between the seemingly simple task it is and the actual complexity that underlies it to make a truly workable system.

BLE doesn’t work for this application. Far too many false results, both positive and negative. Everybody will try to use BLE due to the low cost, long battery life, and that it is built in to your phone (there are already BLE tracking apps, of course). Using BLE for this is like trying to make a car navigator that uses LORAN.

There are 3 fundamental challenges with the UWB social distancing tracking app: battery life, scalability, and body occlusion.

I’ve already touched on the battery life issue. UWB is energy expensive in receive mode and without any infrastructure to synchronize things, you have to do that to hear other tags. One treatment is a large battery and recharge every day as I said above. Another possible treatment is to use BLE as a first contact tactic and then use UWB after the BLE contact is made. This requires some clever protocol development and has substantial complexities not obvious at the start.

Scalability comes from the fact that the possible distance relationships go up by the square of tags in an area. 10 tags is trivial, 100 tags is impossible. The system will have to down select which tags to range to and that becomes a tricky problem without knowing how far away they are. So this will cause false negatives (tags that are close but aren’t ranged to). The users will simply have to accept that the system will be unreliable under conditions of large numbers of people nearby, exactly when you want it to be more reliable. Tuning the algorithm for who to range to will not be easy or trivial, set too low, many false negatives, set too high, air time is overloaded and the system falls down.

Body occlusion comes from where the tag is worn on the body. In most cases, products are focused on some sort of lanyard style tag where the body basically blocks half the directions around you. This will cases false negatives as someone can be close but their tag registers far away due to multipath around the body. If the tag can be mounted on the shoulder or, best yet, top of a head or hard hat, then it works very well. That does not describe most use cases, however.

Spend some time with Google and you can find numerous companies claiming to have UWB based social distancing equipment. They are mostly vaporware, but those who aren’t will all demo well in small settings for short periods of time. You need to find a customer using it at the scale you intend to use it to be sure it will work for you. If you find such a solution, please tell us so I can direct everybody asking for this to somebody who has done it.

Mike Ciholas, President, Ciholas, Inc
3700 Bell Road, Newburgh, IN 47630 USA
[email protected]
+1 812 962 9408

1 Like

Thanks for all your help, we have managed to do it. Like you said there was a lot of underlying complexity to make this seemly simple system work reliably. You were also right about battery life being a bit of an issue, and we are now working ways to save power to increase runtime.

Here is an article and video about our working prototype: link

We are now developing a production unit with built-in contact tracing, reach out if you would like to learn more.

[email protected]

It is a mostly straightforward problem if you install infrastructure, which describes the system in the video.

This system ranges to a set of anchors, position is computed, and then proximity is communicated back to the tag. Such a system can be built with very long tag battery life if you program it carefully, and the system can scale to very large numbers of tracked tags (many thousands) and cover huge areas (several hectares).

But you have to install anchors to cover the area. This is a significant investment in equipment and time.

The challenge is doing this without infrastructure, which is what the majority of the people are aiming for. Then things get very hard for the reasons stated earlier.

Mike Ciholas, President, Ciholas, Inc
3700 Bell Road, Newburgh, IN 47630 USA
[email protected]
+1 812 962 9408

Sorry I didn’t make myself very clear. In the video was a prototype we put together in a couple days, we have since cracked anchor-less detection. Where each tag calculates its distance to other nearby tags without any infrastructure. We put in a factor of safety to account for body occlusion, we have a few tricks to deal with improving battery life and scalability. We have a few clients using our anchor-less prototype currently.

You should provide a video demonstrating that system.

Could it handle the 33 people shown in this photo?

image

That’s 561 possible distance pairs to compute, and just about every person is occluding the people behind them.

If you use signal strength to sub select which UWB ranges to compute, just about everyone here will range to the two guys in front facing them (since they are mostly not occluded and thus have strong signals) and then not range to the guy immediately behind them who is potentially the closest person to you. With the openness of the area, occluded ranges might have very large errors, many meters larger than actual, leading to false negatives.

Even in this relatively small scale situation, just 33 people which could easily be office workers in their cubicles or other common situations, the limitations of capacity and occlusion are severe without infrastructure to help sort it out.

How practical your system is will depend on the answers to these questions:

What is the system capacity in terms of tags in one area?

What is the tag distance update rate at that capacity limit?

What happens when capacity is exceeded?

How big is the battery and how long does it last?

If you can give good answers to these questions, I have potential customers contacting me daily looking for a solution. So far, I have nothing to recommend to them for a no infrastructure system.

There’s no way to do this without causing false positives since you alert too early by the amount of your safety factor when there is good UWB distance measurement.

Occlusion will cause false negatives. The distance a tag will measure when body occluded will depend on what the UWB signal can reflect off of in the environment since the direct path is blocked. In a big empty room, two people 1 meter apart might measure 6 meters apart, for example, when their tags are occluded, if the UWB signal finds something 3 meters away to reflect on. This would be a false negative.

Fundamentally, when the direct path is occluded, there is no way to know what the correction factor is for the distance you actually got. Thus there is no safety factor you can apply that will actually work to provide accurate distances with body occlusions.

Mike Ciholas, President, Ciholas, Inc
3700 Bell Road, Newburgh, IN 47630 USA
[email protected]
+1 812 962 9408

Send us an email at “info at igtimi com” and we can arrange a call to talk about the details. We would love to answer your questions, and a call would be a much more efficient way to communicate.

Not for everyone else reading these forums searching for the company who can truly deliver a no infrastructure reliable social distance contact tracing system using UWB.

If you have this problem solved, you can print money right now, so don’t be bashful about demonstrating you have a truly workable solution.

Mike Ciholas, President, Ciholas, Inc
3700 Bell Road, Newburgh, IN 47630 USA
[email protected]
+1 812 962 9408

1 Like

Hi

I am New to Decawave.

Would be interested to see the response from @igtimi for the questions asked by @ciholas.com?

I wanted to share with you all some of the research we have done with UWB, there may be more. If any one know of resources please let me know.

UWB Chips
* Decawave
* NXP (but it is hard to talk to these guys as they supply Apple and corps)

This seems to be the latest solution for peer to peer social distancing
* https://www.lopos.be/en/
- 99 EURO

* https://www.pozyx.io/
* https://ubisense.com/
* https://www.redpointpositioning.com/
* http://www.forktruckcontrol.com/
* https://kinexon.com/
* https://www.social-distancing-contact-tracing.com/
    - http://www.tsingoal.com
* https://www.sewio.net/
* Carlogic.com
* https://www.eliko.ee/
* https://litumiot.com/
* https://www.infsoft.com/
* Zebra.com
* https://rtloc.com/

UWB Badge company
* Zigpos.de

Very impressive video.
How did you do both RTLS and two way ranging at the same time. Is one an anchor and another a tag? How did the anchor light up?