This simulation shows a large-scale deployment of a mixed robot swarm exploring a shared grid.
The playful cat agents represent lightweight edge devices running compact Q-learning loops to hunt
down their individual objectives. The roaming dog stands in for a higher compute node capable of
executing foundation models sourced from the internet or local stores while continuously refining
the swarm's shared Q-learning state.
- 🐾 Each small cat pushes updates about its discoveries so the dog can aggregate, retrain, and
redistribute smarter policies.
- 🐕 The dog coordinates strategy refinement, injecting large-model insights back into every node
it meets on the grid.
- 📡 Meshtastic-powered LORA links stitch the fleet together, moving telemetry and policy updates
across kilometers on just a few amp-hours.
Watch how the mesh stays resilient as state ripples across the network and agents adapt to their
environment.
Swarm Roster
^.^
Cat Scouts - nimble Q-learning edge bots, low power and dedicated to do specific jobs. Lazy.
/\_/\
( -.-)
( U )
Compot - high compute but expensive dog node, running large models and coordinating the swarm.