Question Please recommend a Router + AP solution for 500+ drones ?

gafonator

Distinguished
Jul 24, 2016
27
0
18,530
Hi, guys,

We've recently pulled off a pretty decent drone show with tens of drones — everything went smoothly using just one (expensive) MI-MO router. Now, we'd like to scale up. We're aiming for hundreds of drones — 500+, actually. But choosing the right WiFi solution for that is slowly becoming a bit of a nightmare. Everyone says something different.

The nutshell version:
We're talking about 500+ clients on a WiFi network, with a large number of concurrent UDP multicast streams running — and we need it all to run as close to perfectly as possible. The distance from the router/AP antennas can reach up to 250–300 meters.

If you’re familiar with such setups, or if you can offer some insight based on your broader networking experience (e.g. which router, brand, technologies, or wireless standards might help, how many APs, rough setup guidance — mesh, repeater, etc.), we’d be grateful for your input.

#RTK, UDP, Multicast, lots of concurrent low-bandwidth streams (several kB per drone), AP mesh
 
Last edited:
Do you actually use multicast. Multicast is very strangely implemented in wifi. It is kinda sent to a broadcast group since sending exactly the same data to multiple device all of which can actually hear the traffic being sent to the other device is kinda silly.

I have not looked at this in years. Multicast was never implemented on the internet where it would be great to reduce the bandwidth of live streamed data. Part of the reason is you can't really encrypt the data and you end up sending separate copies with the unique keys to each end device which defeats the purpose of multicast.

Since all wifi traffic is also now encrypted this makes all the strange issue with multicast on wifi even more complex. In effect it is not actually using multicast it is just sending data via UDP to each device.

In your application a better method would be to use broadcast and let the end devices pick out which messages are for them some where at the application level. You would also encrypt at the application level rather than at the wifi level to keep some silly kid from messing with your stuff. I would assume someone has already though about this. It would make the choice of wifi equipment much simpler and use much less bandwidth.
 
  • Like
Reactions: gafonator
Do you actually use multicast. Multicast is very strangely implemented in wifi. It is kinda sent to a broadcast group since sending exactly the same data to multiple device all of which can actually hear the traffic being sent to the other device is kinda silly.

I have not looked at this in years. Multicast was never implemented on the internet where it would be great to reduce the bandwidth of live streamed data. Part of the reason is you can't really encrypt the data and you end up sending separate copies with the unique keys to each end device which defeats the purpose of multicast.

Since all wifi traffic is also now encrypted this makes all the strange issue with multicast on wifi even more complex. In effect it is not actually using multicast it is just sending data via UDP to each device.

In your application a better method would be to use broadcast and let the end devices pick out which messages are for them some where at the application level. You would also encrypt at the application level rather than at the wifi level to keep some silly kid from messing with your stuff. I would assume someone has already though about this. It would make the choice of wifi equipment much simpler and use much less bandwidth.
Yup. I agree that it is a bit strange. But from what I understood, Multicast is used because as soon as you get many drones at once, the Unicast becomes unmanageable. How it is managed inside the devices themselves, I do not yet understand. Fyi, there are also GPS corrections being sent al the time.

The actual application layer is called RTK, but the whole thing goes somewhat deeper, the NTRIP protocl is as low as Layer 4 (OSI).

OSI LayerRole in RTK contextExamples
Layer 7 – ApplicationApplications that generate or consume RTK correctionsGNSS receivers, RTKLIB, base station software
Layer 6 – PresentationData formatting, message types, encoding (RTCM protocol itself)RTCM 3.x message formats
Layer 4 – TransportEnsures delivery of RTK data streams over networksNTRIP protocol

As for the "let the device choose...", I appreciate you taking time to think about it all, but I don't think we have another option — other than what the applications and control units themselves actually do (Ardupilot and Skybrush). At least not if we don't want to change the whole setup, from hardware to software, control units, etc. I guess...?
 
Last edited:
  • Like
Reactions: gafonator
The largest thing that would be a never buy this product is that it is "cloud managed". It likely can be configured manually but anything "cloud" is a security exposure.

Most the rest of it is your standard marketing guy speak. They can talk about coverage and antenna and everything else but that is only 1/2 the connection. They likely tested between 2 of their units rather than testing to a actual end device. I am sure they did not actually use QAM1024 at those distances. This encoding is extremely dense and wifi signals degrade exponentially with distance so these dense data encodings do not go very far. They of course do not every give any details on how and where they get their numbers from.

So called "seamless" roaming does not really exist since the end device not the network decides where to connect. Now they may load some software on the client, they have in the fine print that you need to use their controller which I assume is some kind of hardware device....or maybe it is a AI cloud based thing 🙁

This does not make this a bad product. Every single company is going to have this marketing speak in their ads.

Bottom line I don't know how you can tell which products would work well for this application. All the actual wifi chips they use are mostly made by mediatek or broadcom. I am not sure if qualcomm only makes end device chips or if they also make chips for router and ap. What this means is there is not a lot of difference between the brands. If you look at the chips they use they will have similar performance.
 
  • Like
Reactions: gafonator
I believe the industry is quite competitive now.

View: https://www.youtube.com/watch?v=MRbBhBAv2dM



Skybrush recommends the RT-AC5300 Asus router, so that is what we used.
But I think he did not mention exactly how many drones or the height he reached,

 
Last edited:
Guys thanks a lot for all the replies. I'll try to sum up the feedback:

- We know how to make a drone show. We already made much bigger shows than the one on the video, at that. But we now aim for 500+ drones. That's a wholly different game, so the video unfortunately doesn't apply all that much.
- The distance is going to be up to 250 - 300m from the AP(s}. - kudos for mentioning the ranges,
- We actually used exactly that one recommended router — RT-AC5300 Asus, however, we feel that it won't be nearly enough for the amount of drones we aim for in the future.
- We used a notebook with appropriate software as a control station + drones OS. - Skybrush, Ardupilot.

If any of you could recommend the AP (mesh) solution together with the router and some insight into whys and hows, that'd be awesome.
 
Last edited:
  • Like
Reactions: gafonator
Your application is very specific and it really depends on how the software functions.
There are 2 issues and likely more that I don't know about since this is not really what wifi was designed to do.

The first issue is most routers can only have so many devices associated. It partially is related to how much data is being send and received but the router must keep track of the mac address and encryption keys for each session so it is also a memory thing. For a office install you generally want less than 30 devices and most AP/routers can handle much more than that. But again how many they can really handle is related to how the application functions and how much communication is done. It the data was really just multicast as indicated in a previous post there really is no session and it would just transmit one stream. I suspect this is not how it functions in your case.

So to fix this first issue in a normal condition you would use multiple AP or the silly "mesh" garbage. Mesh is using a wifi repeater signal between the wifi radios so it is causing even more competing signals. The second problem with "mesh" is it uses single SSID for all the units. There is no way to control where anything connects. All the drones could connect to the same wifi source and you are back to the overload issue stated above.

What might work is if you could use different SSID on the differnet AP. You would then manually load balance the equipment. It is all one large network at a application level but you have seperate wifi networks.

Now this is the simplest case. You would put multiple AP all in the same location running different SSID.

Your other issue will likely be coverage. So now we make it more complex and put in multiple clusters of AP again running multiple SSID but duplicating the main node. Even if we take the simplest case of 1 drone and 2 different AP I can't say how you make that work. There is a wireless roaming protocol generally called 802.11k/v/r. This in theory allows the network to talk to the clients and tell them where the best connection is. It doesn't really work all that well in real life installs but your case is special. The huge problem is not a lot of wifi devices support this protocol. It is fairly common on a router or AP but end device do not generally support it. Some of the latter iphones support it. You would have to research your application.

Since this appears to a be a high dollar project does the manufacture of the drones or the maker of the control software have any suggestion on how this is accomplished.
 
Also...what do your competitors use?

This is obviously not the first mass drone show. Other people/companies have done the same thing....with what?
Tried to ask them as a random internet guy, no response yet, hh.
Tried to research using an AI - no specifics yielded.
I'm at a loss so far.

Obviously, we will experiment and do test flights. However, not spending thousands just to try something out would be really nice.
 
Look for directional WIFI devices. The Ubiquiti Loco M2 would be one example. You could orient them so that the directionally of the coverage is up/down. Maybe setup 4 in a box array. All 4 on the same channel. You could then create three groups on the three clear 2.4 Ghz channels. Most WIFI antennas are NOT designed for vertical coverage.
 
WAVLINK AX3000 Long Range WiFi with 70 Acres of Coverage

View: https://www.youtube.com/watch?v=PEJoxo2s3og
You really have to be suspect of any claims like this.

This is a math and FCC regulation thing. The math is actually complex but you can find many calculators that show how many DB of signal strength you lose at certain distances. The loss increases exponentially. You also have to be careful every 3db loss mean the power is now 1/2 at 9db it is 1/8.

The maximum power allowed is 30db and if you use directional AP you can use 36db on most wifi frequencies.

What you will find is the signal level is extremely low at the common 100meter number ( like -80db) you see for outdoor wifi usage. The only way you get it to work at longer distance is to use directional antenna on the remote end which tends to be impossible for any device that is not in a fixed location.

Almost everything you see talks about 100 meters being the extreme limit for usable wifi that is not point to point.

It would be nice if someone with a degree in RF engineering actually put out youtube videos. Everything you see is people you have little technical background showing stuff that you can't be 100% sure is true and there is a lot of content that actually paid advertising in disguise.
 
I totally agree everything should be wary for wifi claims and very likely could be paid ads. If you are a Amazon Prime member (I believe many Americans already are) , you can at least return the product for free, however.

==

BTW, I don't know how this is possible? No mod ???

He has several videos, but I only watched this one.

View: https://www.youtube.com/watch?v=QJKZIVoHzwk
 
Last edited:
https://www.wavlink.com/en_us/product/WL-WN573HP3.html

This AERIAL HD10 is actually the directional AX 3000 model also from WAVLINK,
the previous video shown model is omni directional.


==

Probably the FCC info (not really sure) for WABLINK AERIAL HD10, part# WL-WN573HP3

official
https://apps.fcc.gov/oetcf/eas/reports/ViewExhibitReport.cfm?mode=Exhibits&RequestTimeout=500&calledFromFrame=N&application_id=AWaF9HbAaQSRNLKzEELMfg==&fcc_id=NZ3-WN0003

unofficial
https://fccid.io/NZ3-WN0003
 
Last edited: