[fho]

One thing that is missing from this simulation is that (at least some) ant species walk straight back to their nest once they found some food. They do this (as far as we know) by accumulating each step they take into what is called the "home vector", effectively integrating the whole path they took. That way they always know which way is "back to the nest". There is a vast amount of papers on the topic. Some (personal) highlights:

If you put the ants on stilts they will follow the same home vector, but because each step on stilts is effectively longer they will overshoot the target: https://www.researchgate.net/figure/Fig4-Odometry-by-stride-...

If the ant is placed close to the nest after gathering food, it will still follow its home vector ... away from the nest: https://royalsocietypublishing.org/doi/abs/10.1098/rspb.2004...

Crazy little buggers these ants.

[Retr0id]

Hah, that reminds me of a robotics competition I took part in at school.

Robots were placed in an arena, and the objective was to gather QR-marked boxes, and bring them back "home".

There were also QR codes placed at known locations around the arena, and the intention was that you could use them for navigation. However, the camera systems were pretty flakey (especially under unpredictable lighting conditions), so we wanted to avoid using them as much as possible. So, we put rotary encoders on our wheels, and integrated the readings to calculate a "home vector" for the return journey, just like you described.

At the time, I wasn't aware that this was an ant-inspired technique - but it was very effective, and we won the competition. Thanks, ants!

[fho]

The crazy thing is how good the ants are at this. Especially considering their brain size.

With robots it works ok, but over time small errors accumulate. After travelling for a while the home error will be off.

In contrast desert ants are able to do this trick after travelling for hundreds of meters.

[withinboredom]

I’m sure the ants have some kind of error correction, and even if it is a one-off robot, you can apply a personal error correction, but it will be specific to that robot. There is no “general” amount of error because every motor/encoder/circuit will provide different levels of error.

[Retr0id]

You can't just "error correct" accumulated random errors. The only way to be accurate is to never lose precision in the first place (which makes the ants especially impressive).

[x0hm]

thants

[eclipticplane]

Bless you ants.

Blants. [0]

[0] https://www.youtube.com/watch?v=gaI6kBVyu00 - Look Around You