Running your model directly from AnyLogic

If you are debugging or unable to export your model (e.g. you don’t have a valid license) you can run your simulation experiment directly from AnyLogic and connect to your python script.

To do so, you should take into account the following considerations:

• In your python script, regardless of the framework that you use, you should ensure that:

  • You add the run_exported_model into the env_config dictionary and set to False.

  • Your framework only launches 1 instance of the environment. For example, in rllib that can be set by defining num_rollout_workers=1 and num_envs_per_worker=1 in rollouts section of AlgorithmConfig (e.g. PPOConfig).

• You execute first your python script and then launch your AnyLogic model. When you run your python script, you will receive a message like this:

  (RolloutWorker pid=6909) 2023-03-17 06:37:36,639 [alpyperl.anylogic.model.connector][    INFO] You can now launch your AnyLogic model! 'ALPypeRLConnector' will handle the connection for you.
  

  After that, you are good to proceed to AnyLogic.

This is an example of how your python script could look like:

from alpyperl.examples.cartpole_v0 import CartPoleEnv
from ray.rllib.algorithms.ppo import PPOConfig


policy = (
    PPOConfig()
    .rollouts(
        num_rollout_workers=1,
        num_envs_per_worker=1,
    )
    .environment(
        CartPoleEnv,
        env_config={
            'run_exported_model': False,
        }
    )
    .build()
)

for _ in range(10):
    result = policy.train()

checkpoint_dir = policy.save("./resources/trained_policies/cartpole_v0")
print(f"Checkpoint saved in directory '{checkpoint_dir}'")

# Close all enviornments
policy.stop()