Confirm Simulations Are Working in AnyLogic - PathmindAI/nativerl GitHub Wiki

Before you export a simulation from AnyLogic to NativeRL, you will need to confirm that your AnyLogic simulation runs without error. If it doesn't run in AnyLogic, RL training will fail.

1. Make sure Pathmind Helper is enabled.

The Enabled checkbox must be selected in the Pathmind Helper properties. Otherwise, no actions will be triggered and no RL training will occur.

2. Make sure you are triggering actions correctly in your model and make sure that your simulation contains an end point.

The event trigger lets NativeRL know when to take the next step, while done defines when a simulation ends.

Running PathmindHelper in debug mode is an efficient way to double-check everything. Run the simulation and toggle the Developer Panel. Check to make sure the data you are feeding to NativeRL is valid.

3. Finally, run a Monte Carlo experiment for 10,000 iterations.

Mechanically, Reinforcement Learning operates similarly to a Monte Carlo experiment. If you are able to run a Monte Carlo experiment in AnyLogic without error for 10,000 iterations, then reinforcement learning training should work fine in NativeRL.

Please be sure to ensure:

• There are no AnyLogic runtime errors.
• Take note of the total steps in your simulation as mentioned above. This will give you a sense of how much training you will be able to fit into a Pathmind run.
• Time how long it takes the Monte Carlo to complete. The longer it takes in AnyLogic, the longer it will take in Pathmind. Keep your simulations as short as possible without degrading the integrity of your model.

If any errors occur, please follow these instructions to resolve the error.