EmEvo

• Library + Set of Experiments for simulating Embodied Evolution of virtual agents.
• Bultiin birth and death rules.
• Enable to simulate evolution of reinforcement learning agents.
  • Only reward evolution is implemented, though.
• Very fast, backed by jax and phyjax2d.

CAUTION

While I want to make it as open as OpenAI gym was, I didn't have enough working time to write documents and stabilize API. So now (July 2024), this is just open sourced for reproducibility of my paper and not very usable for others. Apologies.

Experiments in our ALIFE2024 paper

See alife2024 branch.

Development

Tooling is based on nox. You can compile all requirements by running

nox -s compile

Run examples

Test CircleForaging environment

nox -s smoke

Run tests

nox -s tests

License

Apache LICENSE 2.0