Skip to content

Auto Parameter Suggestion

The Auto Parameter Suggestion feature of the Desktop App uses a genetic algorithm to automatically find optimized parameter combinations. It does this by running multiple simulation runs and evaluating each combination based on a fitness value.

For CoBalance to retrieve this fitness value, the Unity project must provide a component that implements the IGeneticAlgorithmFitnessEvaluator interface.

IGeneticAlgorithmFitnessEvaluator

public interface IGeneticAlgorithmFitnessEvaluator
{
    float CalculateFitness();
}

When SimulationAPI.FinishScenario() is called, CoBalance automatically searches for a component with this interface in the current scene and calls CalculateFitness().

A higher return value means a better parameter combination. The genetic algorithm maximizes this value over generations.

If no such component is present in the scene, this step is silently skipped — simulations without Auto Suggestion therefore work without any changes.

Example

using CoBalance;
using UnityEngine;

public class SimulationFitnessEvaluator : MonoBehaviour, IGeneticAlgorithmFitnessEvaluator
{
    [SerializeField] private GameManager gameManager;

    public float CalculateFitness()
    {
        // Higher score = better balance
        return gameManager.FinalScore;
    }
}

SimulationFitnessEvaluator is its own component in the scene. The GameManager calls SimulationAPI.FinishScenario() — CoBalance finds the evaluator automatically and retrieves the fitness value.

Notes

  • CalculateFitness() is called synchronously, immediately before the simulation run ends
  • The fitness value should be calculated based on the entire simulation run, not just the final state
  • Negative fitness values are allowed, but the genetic algorithm always optimizes toward higher values