Pathmind Helper Debug Mode - PathmindAI/nativerl GitHub Wiki

Before training your RL policy, it's helpful to confirm that the data passed to the NativeRL is valid. This is critical to obtain a well-trained policy.

Printing Debug Information

Inside the PathmindHelper properties, select "Debug Mode" to print to console.

From there, run your AnyLogic simulation and open up the AnyLogic console.

Inside the console, the PathmindHelper will print the following information each time an agent performs an action. Inspect this to troubleshoot errors in your simulation.

Pathmind:
  Step: 1
  Observations:
    Array: {stateChartState[0]=1.0, stateChartState[1]=0.0, stateChartState[2]=0.0}
    Number of Observations: 3
  Metrics: 
    Array: {goalReached=0.0}
    Number of Metrics: 1
  Last Action: [0.0]

Definitions

Step - The step associated with each action. Every time an action is triggered, the step count is incremented.

Observations - Outputs observations at each step. Audit these values to confirm that observations are computed correctly.

Metrics - Output metrics each step. Audit these values to confirm that metrics are computed correctly.

Last Action - The last action taken.