LoRa, LoRaWan, Edge Devices, Opinions

Update here (March 6th, 2021):

After 2 x 100 km drives to find stable Gateways, I have had no luck getting the PortentaH7 Vision Shield (LoRaWan version) working on TTN either V2 theThingsNetwork.com or V3 thethingsindustries.com.

I did however get packets sending and receiving on the helium.com “People Powered HTN Crypto Network” that is big in the States and is getting bigger in Europe. https://explorer.helium.com/coverage

If I fully get it working I will make a video on my Edge Impulse Playlist, but so far the default MKRWAN setup worked for me fairly easily. Helium gives 10,000 data credits to start which relates to bytes. My testing used up 13 credits.

Not really what I wanted, but at least it worked (so far) and it looks like TTN is going towards a pay-to-use model anyway.

Would be very interested if this thread could be turned into a ‘community project’. I’m very new to the TinyML / LoRa ecosystem - playing with some commercial implementations - and would love to get some hands-on (aka hand holding) experience to demonstrate a simple environmental project. Whether that be the ‘sheep tracking’ or ‘bird calls’ TinyML project, for me it’s about learning more about the deployment architecture. Is this in scope for Edge Impulse to set up and moderate? The idea being that we as a community would all go out and get the same board, shield etc. and together would configure the board and run through a tutorial to get a project up and running.

@GlennML Interesting idea, defering to @aurel and @jenny who manage the community.

Just reposting some things that we’ve done:

@janjongboom Thanks for the quick follow up. I’ve gone through these examples and would love to actually be able to demonstrate with my own sandbox. I don’t have the background in embedded tech to fight through some of the issues that @Rocksetta has already run into. To be honest, I get mixed messages about what boards and shields will actually work together. Would love to have one ‘blessed’ configuration that I can work on with others to implement your examples. If that’s the portenta-lorawan project, great! Once we have the sandbox environment, then we can start pushing the value of TinyML & Edge Impulse. IMHO.

Short term: Portenta + LoRa Vision shield indeed as the blessed combination, or the ST IoT Discovery Kit + SX1276 shield with some custom work (https://www.edgeimpulse.com/blog/adding-machine-learning-to-your-lorawan-device).

That community project is an interesting idea! I would definitely work with @Rocksetta on ways we could go about doing this. We could always setup a community GitHub repository that everyone could contribute to - what are your thoughts?

@jenny I would love to become a part of this and contribute as much as I can to the community! :slightly_smiling_face:

@arijit_das_student Thank you! Please let me know if I can be of assist with ya’ll on this community project. With setting up a GitHub repo, etc. Let me know!

Awesome! :slight_smile: It would be very helpful if you can create a GitHub community repo and I will then take on creating some contributions and guides!
(Also would be better if you could add me or @Rocksetta to the GitHub org…so that we might be able to make necessary changes to it instead of forking and making PR’s everytime…what do you think about this ?)

@janjongboom What would be the best way to go about this, create a team in our GitHub org for the ambassadors? Or a new GitHub organization? Or just have Arijit create his own repo?

@arijit_das_student

Just an update here.

I am actually working on a community github but for the Arduino Portenta not specifically Edge Impulse LoRaWan. I am actually doing short zoom meetings Wednesdays 6:00 am PST which is 2:00pm CET (Europe) . Everything will run out of this github

which will become an Arduino library of example code. I have already included an Edge Impulse section. Anyone welcome, link posted on the site just before the meeting each week.

As far as The Portenta LoRaWan Vision Shield an update about Point-to-Point LoRa is here https://github.com/arduino-libraries/MKRWAN/issues/87#issuecomment-805028506

I also have the Helium network working on setting up the Portenta.

3 Likes

I had the opportunity to sit with @Rocksetta yesterday morning and have also started reviewing some of his @Rocksetta youtube videos) related to TinyML and working with the Ardunio platform. At a high level, I think this is the type of approach that I - as a total newbie to both TinyML and micro-tech - am looking for. If I understand correctly, Edge Impulse could use it’s enterprise version to foster team collaboration on a community project, maintain a related GitHub project/discussion group and ideally, host a weekly(?) zoom call with the community project members to make sure everyone is moving along in tandem? Kind of a Meetup meets EdX model. To be honest, just trying to implement one of your sandbox examples would be a good starting point. The less I have to fight with the hardware environment, the sooner I can focus on TinyML and look to extend it into separate verticals (IMHO). Thx.

It was great talking to you yesterday! Looking forward talking more about your project. :slight_smile:

:+1: Thx! Decided to jump into the deep end and just ordered the Portenta and Vision Shield (LoRa). I really hope this doesn’t become a very expensive paper weight. “sigh

1 Like

Update here about LoRaWan and the Portenta Vision Shield.

I have point-to-point LoRa working and I need to work a bit more with it, the breakout board having Ethernet and the Shield having LoRa is an interesting combination.

With TheThingsNetwork (TTN) in some flux moving from V2 to V3 and not many gateways nearby, and having some issues setting up my Gateways while things are in transition. I will put this off or try setting up the Open Source version of TTN, which some people are having success in North America.

So I have been working on connecting to the Helium Network.

Well yesterday I finally connected and downloaded data! :grinning: I used Adafruit.io MQTT integration with CayenneLPP to JSON parse the data. Unfortunately my Portenta method is way to complex, so I will try to make things a bit simpler.