Woven robot swarm weaves at Algomech (with a little help from their friends)
A few pictures and videos from last weekend’s debut of the Penelopean weaving robots at the Algomech festival in Sheffield. After a frantic couple of weeks in mass production, a swarm of 8 robots and a may pole invaded the winter gardens, wove and un-wove braids to the sounds of livecoded beats and ancient Greek poetry.
They were set up in a simple choreography, with two groups of 4 robots taking it in turns to rotate in opposite directions around the maypole to create an overlapping twist pattern. This was orchestrated by a Raspberry Pi with each robot reporting back it’s telemetry using radio messages. We tried different walk speeds depending what else was going on, but they took approximately half an hour to create a full braid and another half hour to undo it.
This is a short update – but this is what we have learned:
- The robots survived in total 6 hours of continuous running with lots of handling by children.
- The wooden legs and a new woven construction resulted in a very consistent behaviour amongst the swarm, with only one being problematic.
- They are highly approachable – people generally (particularly, but not exclusively children) want to pick them up and help them, try changing their direction and so altering the weave.
- Having one that is consistently falling over encourages people to get involved.
- A dance pattern where they take turns to move, and therefore spend half their time still means there is a chance for people to pick them up.
- The musical synchronisation causes the servo motors to periodically slightly twitch, meaning that even when not moving they indicate that they are active. This is particularly effective when tied via the ribbon to the may pole, as it amplifies the effect.
- Most of the work on the software and electronics was simply getting the radio messaging set up properly to be consistent and fault tolerant when broadcasting to lots of robots. It had to deal with quite a few dropped messages, and problems with the robots resetting themselves sometimes when handled, but it passed this test.
Also check out the penelopean robotics github to keep up to date with developments.