Running your simulation model as an AnyLogic `CustomExperiment`

The CustomExperiment is a specific AnyLogic experiment type, enabling developers to generate and control experiments solely through code, without utilizing a graphical interface or predefined settings. For instance, in a Monte-Carlo experiment, users can define custom expressions for the simulation model parameters for each iteration and the output value to display in the results histogram.

This experiment type is commonly employed by developers to run the model without any visual elements (no simulation window will be launched), enabling interaction with the model exclusively through the Console. This approach is particularly advantageous when the primary focus is on performance or on running as many iterations as possible in a minimal time frame. This is facilitated by the absence of visual element rendering.

Warning

However, ALPypeRL requires you to notify the ALPypeRLConnector that it is running under a CustomExperiment and pass the required information (which include the experiment command line arguments).

At the moment, ALPypeRL is unable to collect this information by itself.

Here’s a guide on how to proceed in case of the CartPole-v0 example:

// Create Engine, initialize random number generator:
Engine engine = createEngine();
engine.setTimeUnit( SECOND );
// Fixed seed (reproducible simulation runs)
engine.getDefaultRandomGenerator().setSeed( System.currentTimeMillis() );
engine.setStartTime( 0.0 );
engine.setStartDate( toDate( 2023, FEBRUARY, 28, 16, 0, 0 ) );
// Set stop time:
engine.setStopTime( 2000.0 );
// Create new root object:
Main root = new Main( engine, null, null );
// Setup parameters of root object here
root.setParametersToDefaultValues();
root.set_rlMode(RLMode.TRAIN);
// Set ALPypeRLConnector parameter values
root.alPypeRLConnector.set_connect(true);
// Set 'Action' and 'Observation' spaces
root.alPypeRLConnector.set_setSpaces(true);
root.alPypeRLConnector.set_observationSpace(
    ObservationSpace.init()
        .add(GymSpaces.box(-2 * 2.4, 2 * 2.4))
        .add(GymSpaces.Box.unbounded())
        .add(GymSpaces.box(-2 * 12 * 2 * PI / 360.0, 2 * 12 * 2 * PI / 360.0))
        .add(GymSpaces.Box.unbounded())
        .build()
);
root.alPypeRLConnector.set_actionSpace(
    ActionSpace.init()
    .add(GymSpaces.box(-1.0, 1.0))
    .build()
);
// Set unique simulation runs
root.alPypeRLConnector.set_randomnessMode(RandomnessMode.UNIQUE_RUNS);
// Notify ALPypeRLConnector is running under a custom experiment
root.alPypeRLConnector.isCustomExperiment(getCommandLineArguments());
// Prepare Engine for simulation:
engine.start( root );

Note

If you need to, do remember to set the rlMode as RLMode.TRAIN as shown in the example.

This is required in the package examples as the mode is set via an additional parameter. If you define this parameter directly at ALPypeRLConnector then you do not need to specify its value in the code.

If you want to test a CustomExperiment or use it as a base model for your project, you can find it at:

The experiment has been created and stored inside the .alp file for the CartPole-v0 example at ./alpyperl/examples/cartpole_v0/CartPole_v0.
There is also an exported version located at ./resources/exported_models/cartpole_v0_custom_experiment. Remember to point to this folder in your train.py script env_config and exported_model_loc when defining your environment.

Running your simulation model as an AnyLogic CustomExperiment

Running your simulation model as an AnyLogic `CustomExperiment`