Rovers in chicken sheds

IMG_20211117_130249

Hi there, Szymon from Leo Rover.
2 years ago we partnered with FLOX.ai and develop robots to be used in chicken sheds to make the short life of the chicken much more bearable.

More here: https://www.youtube.com/watch?v=Uv9FXqCaD4U

Here’s a question: do you know good a quality indoor positioning systems that we could use of-the-shelf?

We’re experimenting with different approached already and any suggestion would be great! The use case is quite specific because we’re talking animals here and for ex. Marvelmind’s ultrasound based beacons might be annoying for the chickens themselves. (they are annoying for me and other developers already :face_with_thermometer: )

Thanks!

7 Likes

Hi Szymon,

very interesting project.

How accurate positioning is required? Is 10 - 20 cm enough? If yes, then a UWB based system (similar to Apple AirTags) could be used, e.g. based on this family of chips: DW3120 - Qorvo

Kind regards,
Gergely

Wow, this post brings back memories. I used to work on deploying a rover to chicken sheds too. Here are some things I’ve learned in no particular order:

  • The material on the ground will change with time. For some platforms it will be a challenge to drive through it as more and more poop is stuck to the wheels
  • Young chickens are fearless, they will often peck the rover. Older ones will mostly keep their distance but not always
  • Some chickens will refuse to move away and clear the path for you
  • Once your robot is in the shed, it’s a hassle to take it out - you need to thoroughly clean it with some alcohol based solutions. Ingress protection is highly recommended to make things easier
  • The environment is hot and humid, moving the robot from a cooler place into the shed can cause some condensation in some bits
  • Chicken cultivation is brutal. Broiler chickens that are used for meat grow really fast, to the point that their bones can’t support their legs. Even though they mix antibiotics with the water, there will be some chickens dying and as you are testing your robot you might find yourself standing next to a pile of dead ones ready to be taken away. And these are okish conditions for these birds. Cages are much worse
  • On the sad reality of dead chickens, you probably should not drive over them with your rover if you are able to detect them
  • On the antibiotics: one day I arrived feeling ill to test in the shed, and after half a day there I was no longer sick. I was always wondering if the antibiotics make it into the air and that cured me

This was one of these experiences that made me buy only free range meat and eggs. Fun fact: in EU free range means that the chickens have access to an open-air run. It doesn’t mean that they will actually use it and be all happy.

Sorry for getting quite a bit off topic here, but I thought someone might find this rambling interesting.

3 Likes

I think the solution by pozyx also would work. They also use UWB technology.

@kisg Yes, I believe 10-20cm would be ok. We’ll check that, thanks.

@msadowski Interesting insight. Actually we’ve already tested those in several occasions. Fortunately the sheds we work with are already ‘organic’ ‘free range’ etc. as those are more happy to introduce FLOX’s systems. As I’ve been to Menchine (shed) everything looked much nicer than I expected :slight_smile: and even there, as you mentioned - free range chickens don’t necesarily use their freedom

@JeremyLebon Sure we’ll buy everthing from the market and validate. Thanks!

When I tested UWB in the early days (around ~2016-2017) we had quite a few issues with metallic walls reflecting the signals and adding up noise. I imagine the things improved significantly since then, but that’s one of the things I would check early.

The FLOX system mounts some overhead cameras, right? Could you use them for localization? I imagine that with the right calibration you should be able to transform the pixels to global coordinates?

Good info.

Yes FLOX uses the cameras, but at the moment their lenses and lack of exact positioning (for ex. after cleaning) makes it super hard to get this centimeter accuracy. Especially there are mutliple cameras with their images stitched.

We’re just looking for something OTS with hope to improve the stitching in the future.

Hi @Szymon_Dzwonczyk,

you probably already thought about this, but one option could be to place IR led markers with unique blinking patterns in the FoV of the cameras at well-known positions which could be used to automatically recalibrate the cameras even if their position is moved.

Then you could use the same solution with blinking IR LED patterns on the top of the rovers themselves.

The same solution was used by the Oculus Rift DK2, that was reverse engineered:

Kind regards,
Gergely