N Variations on a Theme

Back in pre-history I had a job doing robotics research which lead me to make some small autonomous vehicles that, thanks to my friend Tori's support in showing something at her Dactyl space in Soho, developed a sequence of motions for themselves -- taught themselves a dance step.

Here's the whole presentation from a time when the internet sort of worked for video, which it now does in a completely differently-abled way (I think the video files are still there but they are in Quicktime which seems to have been disappeared):

ROBOCAR Collective

tldr; Each robot had a set of motions it could do -- forward, backward, and turns. It would pick a partner and each would select a motion and then adjudicate who's movement to execute.  This would make that motion more likely to be selected, by both partners, in future situations. Then they would pick a new partner and repeat the exercise until, eventually, all five of the vehicles would be performing the same sequence.  This was programmed using a Stochastic Finite State Automata, which is a fine state of affairs if you are not jargon enabled.

Unfortunately, I (they) had no way of knowing where each vehicle was and which way it was facing. So the net result still looked a bit of a jumble. But. Here is the U-Tub money shot from when "they" all happened to synchronize (at https://youtu.be/slfprpe848s):

I tried to fix the position insensitivity a couple times but eventually lost interest due to many of my pre-existing conditions.

Funny(?) aside. Now, ten or so years later, it appears that Urs Fischer has (hired folks to) solve this problem using a set of office chairs, as displayed at the Gagosian Gallery in Chelsea:

Gagosian -- Fischer -- Chairs