Collision in group poll


As i understand tags will receive data for occupied time slots per beacon in the beacon message. It will then check if there are any slots in which all 3 or 4 beacons are free. But it will reserve it only in the group poll slot.

As the dwm1001 system description document part 7 states it is theoretically possible to use up to 9000 tags on 60 seconds interval in the same area. What stops all these tags to all request the first TWR slot in the first group poll?
Will this create a never ending stream of collisions? How is this handled?

Thanks in advance,
Kind regards

Hi JK,

When a tag request TWR slot, it will wait for acknowledgement from at least 3 anchors (it is asking permission to 4 anchors).

If it does not receive 3 acknowledgement, then the tag will drop the twr slot and try again in a randomly determined time. This is ensuring the system will converge as all tags will request slots at different time.

Let me know if it’s clear,

Hi Yves,

Thanks that’s great! Any idea or insights in how long it would take for approximatly 5000 tags to get a timeslot?(i assume this time will be random within 1 to 60 seconds).
Is there also a random function for TWR slot select? or will it always request the first slot if this is not taken?

Your help is appreciated!
Kind regards,

Hi Jk, I would say between 33 sec and 120 sec .

At best 15 tags can get a slot per super-frame, and there are 10 super-frames per second so it is roughly min 33 second for each tag to get a slot. That’s an ideal case, I would suspect there would be collision most likely. The system will probably stabilize after 60 sec, but I am not too sure about how long it will take.

Please note to support 5000 tags tags will have to use to reduce update rate.

The tag receives the list of free slots from anchors and will try to reserve the first available slot, and so on until it get a slot confirmed.

UPDATE: I’ve been doing some experiments, and I have to say I was wrong, I would think the system will take longer to stabilize. Unfortunately I can quite tell as it depends on how collision will be resolved. Keep in mind PANS network is not built for very large density/size systems. It may be used for them but I would recommend you to perform field testing to stress the system and observe best performance.


Hi Yves,

Please correct me if i am wrong.

5000 tags / 10 superframes * 60 seconds = 7/8 tags per superframe. if all the tags would in order check if they can range with at least 3 anchors as you mentioned.
Wouldn’t the chances of arriving at all 3 anchors as firtst of the 8 tags be 1/8 * 1/8 * 1/8 = 0.001 which is 0.1%

Doesn’t this mean that only 0.1% of the 5000 tags would stabelize in the first minute?

any place where i can find more info on this? I already checked DWM1001 system overview section 4.6 and section 4.7.

Hi JK,

We don’t have more information about that i’m afraid. I’ll have to double check with the r&d team.

Also you are kind of assuming all tags will try to join the network at the exact same time from the same physical place and I don’t think this is very likely to happen.

Please keep in mind PANS is initially developed for small to medium RTLS system. For a very large number of tags you may want to look into TDoA.

Thank you,

Hi Yves,

Thank you, i would like to know more about this topic.

Topic dwm1001 BCNx, SVCx and TWR slot x occpy time states that the group poll takes 500 us.
quick calculation states that it would take 31.7 us for the group poll data to transfer. I understand that it takes some time to get ready to reply .

But this big diffrence makes me wonder if there is a random function regarding time in which to send data in the group poll?

i really hope you can give me an answer on this question, since it would be essential for my project.

Kind regards,

Hi JK,

I think one of the statement is not very accurate. The slot during which group poll is sent is 500us, but the group poll message will be shorter. I don’t think there is randomization of time to sent the gorup poll within these 500us.


Hi Yves,

some tests i did confirmed that indeed always the first TWR slot is chosen if possible. For my understanding this means that 50 tags would stabilize per minute as stated above.

I know its a long shot but may it be possible that i get a firmware version in which choosing the svn slot will happen randomly within all slots that are not occupied? This would make the tags that stabilize per minute much higher.

I will then proceed to use the Decawave TWR package for my 100000+ tag system.

kind regards,

Hi JK,

At the moment we cannot perform customization of PANS on a customer request base sorry.

It sounds like you have plan for very large system that will need extensive hardware and software support. Maybe you can contact the Decawave sales channel to pursue the discussion. Depending on your plan and schedule, they will help providing the best support.