Inquiry About Reinforcement Learning Integration with APSIM

[pathakprecisionag]

Greetings,

I hope this request aligns with the interests of fellow contributors. I regularly use APSIM to simulate crop production for "what-if" scenarios, and I’m exploring the possibility of integrating it with reinforcement learning algorithms.

Is there currently a module or package available that supports this kind of integration, similar to how DSSAT has the gym-dssat environment? If not, could anyone point me toward any existing work or resources that might help facilitate this connection (e.g., using APSIM as an environment for reinforcement learning)?

Any insights or guidance would be greatly appreciated!

[peter-devoil]

"pan dot zhao at ua edu" had a student working on an ApsimNG/RL application, I think it's still a work in progress. His method uses a ZMQ interface in NG - which is mostly complete and allows you to write a manager component that talks with an RL algorithm - fielding requests about the state of the system, and responding to interventions like sowing or fertilising crops.