library implementing barebones gbp for use with factor graphs. Based on work
- https://colab.research.google.com/drive/1-nrE95X4UC9FBLR0-cTnsIP_XhA_PZKW
- https://gaussianbp.github.io/
- only a goal layer.
- click to assign a new goal state
$x^G$ . - agents can always communicate with each other.
- they optimize to push their goal state a minimum distance apart from the other agents' goal state.
-> done see examples/diverge_from_line.py
- now do it in 2d
- make an event loop
- [ ]
- implement MPC.
- map landscape for mapping discrete action methods.
- make plan for how to operate over a navigation graph instead of over a continuous space.
- go from goal layer to task layer.
- mpc is also just pushing nodes apart that lay too close together.
- information layer.
- so