Building in-cab control for multiple OWLs

[geezacoleman]

It's be a while coming (first raised in Issue #5 ) but I've started work on an MQTT protocol for communication between multiple Raspberry Pis mounted on a boom.

The architecture would be:

• One Raspberry Pi acting as a MQTT broker and also running an OWL GUI that subscribes to relevant topics
• The Pis on the boom would publish information on:
  • each detection/weed location (assuming GPS integration)
  • status - i.e. any errors or otherwise
  • subscribe to an 'operational' topic for stop/start/blanket spray etc. instructions
  • subscribe to 'parameters' topic for sensitivity/changes in algorithm
• cab-mounted Pi would then display when nozzles are on and allow user to interact and adjust performance remotely with a screen/button interface.

Have started on this in the mqtt-comms branch with a basic PyQt5 UI. All ideas/contributions/suggestions are welcome!

[Minorplanets76]

AgOpenGPS has migrated to UDP via ethernet with messages being sent via PGN's. The upgrade has done wonders for system response and reliability. CANBUS would be another possibility given its use in many tractor and spray technologies. There is a branch looking at rate control as well, and ESP based wifi which may provide some inspiration.

[EchelonMirror]

I moved to Wifi mostly for scalability, and some future plans with error detection, calibration via tablet etc. Ethernet switches that are IP67 compliant are prohibitively expensive. For small scale systems, an 8-12 port switch will do the job if you put it in a housing that is compliant.

For large scale systems, I've had issues with EMI caused from tractors and large noisy power sources. geezacoleman, someone in your fam might be of help here to avoid some of the issues of larger hardware.

[sunlite71]

Dear Minorplanets76, Just curious have you tested Owl with Esp32 Cam?
Your earliest reply will be highly appreciated :-)

thanks

[ahogan27]

Where did this line of thought end up @geezacoleman? To Wifi or not to Wifi?

[geezacoleman]

Though fairly uninformed about wifi, I'm still of the belief that the hard wiring of modules is more reliable. Spoke to DeepAgro (Argentine spot spraying company) and they mentioned wifi had caused some trouble. Ethernet would allow video streaming and MQTT commands. But, a CAN system seems to be most common, particularly if we want to communicate with third part screens or robots (e.g. Agrointelli or SwarmFarm), I have also heard bad things about CAN and reliability.

So I think end result will be ethernet and MQTT - it's most future proof, easily done and could potentially be switched with wifi in the future if needed.

[EchelonMirror]

CANBUS is ancient and unreliable. ISOBUS is heading in the same direction since the standard was developed in the 90s.

Ethernet is hard to beat on reliability, though a lot of people will complain about wiring up extensive amounts of looms and the question of where you put the switch. Since it's going to need power and easy access for cables, in addition to IP64+ casing.

Wifi can work if you reduce the number of units you have on the boom. I've been testing with a number of routers from Teltonika (RUT655) on a 36m boom and it's mostly stood up to scrutiny - I must note this is with 2-3m spacing with custom antennas and an Azurewave AW-CB375NF. Key is to make sure your devices have a rough LOS (line-of-sight) to your router. I'm using MQTT commands on that system.

Basically, there is a core that acts a web server and MQTT broker. The goal was to be able to use a web browser as the control system through a cheap android tablet. Being purely software, it allows you to roll up new features with a single software update of the core.

Do the folks at DeepAgro use real-time streaming?