private ApplicationEngine(Size worldCanvasSize, Size brainCanvasSize, Size graphCanvasSize)
{
_geneticEvolution = new GeneticEvolution(ApplicationSettings.Random, ApplicationSettings.MutationRate, ApplicationSettings.CrossOverRate);
_agentsColor = new List <Color>();
agents = new List <Agent>();
food = new List <Food>();
badFood = new List <Food>();
bitmapWorld = new Bitmap(worldCanvasSize.Width, worldCanvasSize.Height);
graphicsWorld = Graphics.FromImage(bitmapWorld);
bitmapBrain = new Bitmap(brainCanvasSize.Width, brainCanvasSize.Height);
graphicsBrain = Graphics.FromImage(bitmapBrain);
bitmapGraph = new Bitmap(graphCanvasSize.Width, graphCanvasSize.Height);
graphicsGraph = Graphics.FromImage(bitmapGraph);
typeOneEats.Add(SpecieType.Plant);
typeTwoEats.Add(SpecieType.Herbivore);
_agentsColor.Add(Color.Green);
_agentsColor.Add(Color.Red);
_agentsColor.Add(Color.Blue);
for (int i = 0; i < ApplicationSettings.NumberOfAgentsTypeOne; i++)
{
Agent a = GenerateAgent(SpecieType.Herbivore, false, null);
if (a != null)
{
agents.Add(a);
}
}
for (int i = 0; i < ApplicationSettings.NumberOfAgentsTypeTwo; i++)
{
Agent a = GenerateAgent(SpecieType.Carnivore, false, null);
if (a != null)
{
agents.Add(a);
}
}
for (int i = 0; i < ApplicationSettings.FoodOnScreen; i++)
{
food.Add(CreateNewFood(colorFood, 50));
}
for (int i = 0; i < ApplicationSettings.BadFoodOnScreen; i++)
{
badFood.Add(CreateNewFood(colorBadFood, -50));
}
evolutionGraph = new Graph(200, 500, 3, _agentsColor);
evolutionGraph.AddPoint(0, 0);
evolutionGraph.AddPoint(1, 0);
evolutionGraph.AddPoint(2, 0);
}
public Evolution(int maxGenerations = 100)
{
_maxGenerations = maxGenerations;
var options = new EvolutionOptions
{
PopulationSize = 100,
MaxNumberValue = 100,
CollectionSize = 4,
Crossover = CrossoverTypes.Slice,
Mutation = MutationTypes.Addition,
};
_geneticEvolution = new GeneticEvolution <SnakeGA>(options);
}
private ApplicationEngine(Size worldCanvasSize, Size brainCanvasSize, Size graphCanvasSize)
{
ticksIntoGeneration = 0;
ticksForGeneration = 10;
timeWhenApplicationStarted = DateTime.Now;
_geneticEvolution = new GeneticEvolution(ApplicationSettings.Random, ApplicationSettings.MutationRate, ApplicationSettings.CrossOverRate);
_agentsColor = new List <Color>();
agents = new List <Agent>();
food = new List <Food>();
badFood = new List <Food>();
bitmapWorld = new Bitmap(worldCanvasSize.Width, worldCanvasSize.Height);
graphicsWorld = Graphics.FromImage(bitmapWorld);
bitmapBrain = new Bitmap(brainCanvasSize.Width, brainCanvasSize.Height);
graphicsBrain = Graphics.FromImage(bitmapBrain);
bitmapGraph = new Bitmap(graphCanvasSize.Width, graphCanvasSize.Height);
graphicsGraph = Graphics.FromImage(bitmapGraph);
typeOneEats.Add(SpecieType.Plant);
typeTwoEats.Add(SpecieType.Herbivore);
_agentsColor.Add(Color.Red);
_agentsColor.Add(Color.Blue);
for (int i = 0; i < ApplicationSettings.NumberOfAgentsTypeOne; i++)
{
GenerateAgent(SpecieType.Herbivore, false);
}
for (int i = 0; i < ApplicationSettings.NumberOfAgentsTypeTwo; i++)
{
GenerateAgent(SpecieType.Carnivore, false);
}
for (int i = 0; i < ApplicationSettings.FoodOnScreen; i++)
{
GenerateFood();
}
for (int i = 0; i < ApplicationSettings.BadFoodOnScreen; i++)
{
GenerateBadFood();
}
evolutionGraph = new Graph(200, 500, 2, _agentsColor);
map = new Map(ApplicationSettings.MapSize, ApplicationSettings.CellSize, ApplicationSettings.Random);
}
protected void Evolve <T>(int generationsCount = 50, EvolutionOptions options = null, Func <EvolutionResult <T>, bool> stopCondition = null) where T : class, IEvolutionaryIndividual, new()
{
var geneticEvolution = new GeneticEvolution <T>(options ?? EvolutionOptions.Default);
geneticEvolution.EvolveUntil(stopCondition ?? (r => r.Generation.Number == generationsCount), result =>
{
_output.WriteLine($"Gen. : #{geneticEvolution.CurrentGeneration.Number}");
_output.WriteLine($"Best : {result.BestIndividual}");
_output.WriteLine($"Avg : {result.AverageIndividual}");
_output.WriteLine($"Worst: {result.WorstIndividual}");
_output.WriteLine($"------------------------------");
_output.WriteLine($"Avg.Fitness: {result.AverageFitness}");
_output.WriteLine("");
return(result.Generation.Number < 500); // safe stop
});
}
//Subscribed to PlotStart, which is queued in UIReflect after UI updates (which is queued StartEvolve)
private void EvolvePopulation(object sender, EventArgs e)
{
TimeLogger.LogStart();
float averageFitness = 0.0f;
List <Bed> evolvedBeds = new List <Bed>();
//Actual Evolve process
for (int g = 0; g < givenIterations; g++)
{
//New beds get inited from rand conditions for the first gen
//Should ideally ahve already been done from Start -> ResetPopulation, just a safety measure
if (!initialised)
{
newBeds = CreateBeds();
}
//The previous List gets emptied. This data is now stored in newBeds after the last generation
if (initialised)
{
}
/*for (int i = 0; i < beds.Count; i++)
* {
* beds[i].ComputeFitness();
* }*/
if (evolvedBeds == null)
{
//Debug.Log("Evolved null");
}
if (newBeds == null)
{
//Debug.Log("new beds null");
}
//Use the GA to evolve our data
evolvedBeds = GeneticEvolution.Evolve(newBeds, newBeds.Count, testbench.GetUsedMatingPoolSelector(),
testbench.GetUsedBreedingPairSelector(),
testbench.GetUsedCrossoverOperator(),
testbench.GetUsedMutator());
for (int i = 0; i < evolvedBeds.Count; i++)
{
evolvedBeds[i].LoadGardenAndPatches(this, patches);
}
generationNumber++;
//Debug.Log("Unsorted List " + beds[0].Fitness);
Bed[] bedArray = evolvedBeds.ToArray();
TimSort <Bed> .sort(bedArray, new GeneticEntityComparable());
/* for (int i = 0; i < bedArray.Length; i++)
* {
* Debug.Log("Sorted post-gen i: " + i + " Fitness: " + bedArray[i].Fitness);
* }*/
//Debug.Log("Top of Array: " + bedArray[0].Fitness);
evolvedBeds = new List <Bed>(bedArray);
//Trimming the solutions
//Debug.Log("Evolved Beds pre-trim - size: " + evolvedBeds.Count);
//Certain Pair Selectors create larger numbers of solutions than the intial set of parents.
//In this case, we require as many solutions (beds) as we put in and therefore remove the excess
//after they've been fitness sorted
if (evolvedBeds.Count > numberOfBeds)
{
evolvedBeds.RemoveRange(numberOfBeds - 1, (evolvedBeds.Count - numberOfBeds));
}
//Debug.Log("Evolved Beds post-trim - size: " + evolvedBeds.Count);
averageFitness = 0.0f;
for (int i = 0; i < evolvedBeds.Count; i++)
{
averageFitness += evolvedBeds[i].Fitness;
// Debug.Log("Beds [" + i + "]:" + beds[i].Fitness);
}
averageFitness /= evolvedBeds.Count;
//Clear out the starting beds
newBeds.Clear();
//Save the evolved beds into a new starting List to be used for the next generation
for (int i = 0; i < evolvedBeds.Count; i++)
{
newBeds.Add(evolvedBeds[i]);
}
//Clear out the evolved one
evolvedBeds.Clear();
//Debug.Log("End of gen loop - eb size: " + evolvedBeds.Count + " | nb: " + newBeds.Count);
}
TimeLogger.LogEnd();
EventQueue.QueueEvent(EventQueue.EventType.Plot_End, this, new TimeArgs(Time.realtimeSinceStartup));
//We use newBeds not evolvedBeds here since at the end of the gen, new holds the next (or last) generation's data
//and evolved is already cleared for the next gen
EventQueue.QueueEvent(EventQueue.EventType.BroadcastGeneration, this, new GenerationArgs <Bed>(newBeds, newBeds[0].Fitness, averageFitness, generationNumber, testingBase));
}
