private Particle[] CreateSwarmOfSolutions()
{
try
{
var swarm = new Particle[Config.SwarmSize];
var radomDouble = new RandomDouble();
var braninRcos = new BraninRcos();
for (var particleIterationNumber = 0;
particleIterationNumber < Config.NumberOfIterations;
particleIterationNumber++)
{
// step 1 - create initial swarm of particles
if (particleIterationNumber == 0)
{
for (var particleNumber = 0; particleNumber < swarm.Length; particleNumber++)
{
double randomX1;
double randomX2;
do
{
randomX1 = radomDouble.GetRandomNumber(Config.MinX1, Config.MaxX1);
randomX2 = radomDouble.GetRandomNumber(Config.MinX2, Config.MaxX2);
} while (particleNumber != 0 &&
Math.Abs(randomX1 - swarm[particleNumber - 1].CurrentPosition[0]) < 0.000000000000001 &&
Math.Abs(randomX2 - swarm[particleNumber - 1].CurrentPosition[1]) < 0.000000000000001);
var initialVelocityX1 = 0.0; // first iteration = velocity = 0.0
var initialVelocityX2 = 0.0; // first iteration = velocity = 0.0
var currentCost = braninRcos.BraninRcosObjectiveFunction(randomX1, randomX2);
var particleOfFirstIteration = new Particle
{
ParticleId = particleNumber + 1,
CurrentPosition = new[] { randomX1, randomX2 },
CurrentVelocity = new[] { initialVelocityX1, initialVelocityX2 },
PersonalBest = new[] { randomX1, randomX2 }, // first iteration = personal best = current position
Cost = currentCost
};
// update global best if paricule performs better than previouse global best
if (currentCost < Particle.GlobalBestCost)
{
Particle.GlobalBestCost = currentCost;
Particle.GlobalBestPosition = particleOfFirstIteration.CurrentPosition;
}
swarm[particleNumber] = particleOfFirstIteration;
}
}
else
{
// 2nd iteration onwards
for (var particleNumber = 0; particleNumber < swarm.Length; particleNumber++)
{
var currentParticlesCost = braninRcos.BraninRcosObjectiveFunction(
swarm[particleNumber].CurrentPosition[0], swarm[particleNumber].CurrentPosition[1]);
// move to a new position if the current particle is not global best solution found so far
if (currentParticlesCost > Particle.GlobalBestCost)
{
var newVelocityAndPosition = MoveToNewPostionWithInRange(swarm[particleNumber].CurrentPosition[0],
swarm[particleNumber].CurrentVelocity[0], swarm[particleNumber].PersonalBest[0],
Particle.GlobalBestPosition[0], Config.MinX1, Config.MaxX1);
var velocityX1 = newVelocityAndPosition[0];
var randomX1 = newVelocityAndPosition[1];
newVelocityAndPosition = MoveToNewPostionWithInRange(swarm[particleNumber].CurrentPosition[1],
swarm[particleNumber].CurrentVelocity[1], swarm[particleNumber].PersonalBest[1],
Particle.GlobalBestPosition[1], Config.MinX2, Config.MaxX2);
var velocityX2 = newVelocityAndPosition[0];
var randomX2 = newVelocityAndPosition[1];
//var velocityX2 = CalculateVelocity(swarm[particleNumber].CurrentPosition[1],
// swarm[particleNumber].CurrentVelocity[1], swarm[particleNumber].PersonalBest[0],
// Particle.GlobalBestPosition[0]);
//var randomX2 = CalculatePosition(swarm[particleNumber].CurrentPosition[1], velocityX2);
var currentCost = braninRcos.BraninRcosObjectiveFunction(randomX1, randomX2);
// update particle
swarm[particleNumber].CurrentPosition = new[] { randomX1, randomX2 };
swarm[particleNumber].CurrentVelocity = new[] { velocityX1, velocityX2 };
swarm[particleNumber].Cost = currentCost;
// update personal best
var personalBestCost = braninRcos.BraninRcosObjectiveFunction(
swarm[particleNumber].PersonalBest[0], swarm[particleNumber].PersonalBest[1]);
if (currentCost < personalBestCost)
{
swarm[particleNumber].PersonalBest = new[] { randomX1, randomX2 };
}
// update global best if paricule performs better than previouse global best
if (currentCost < Particle.GlobalBestCost)
{
Particle.GlobalBestCost = currentCost;
Particle.GlobalBestPosition = swarm[particleNumber].CurrentPosition;
}
}
Config.WInertiaWeight = Config.WInertiaWeight * Config.WDamping; // with each iteration w values decreases
}
}
}
return(swarm);
}
catch (Exception exe)
{
Console.WriteLine(exe);
throw;
}
}
