Particle_With_LS.py

from Pallets import Pallets
import random

class Particle_With_LS:
    """
    A variant of the Particle class. Represents a single particle in the PSOwLSI algorithm.
    """

    INERTIAL_COEFFICIENT = 0.721
    COGNITIVE_COEFFICIENT = 1.1193
    SOCIAL_COEFFICIENT = 1.1193

    def __init__(self, pallet_problem : Pallets, initial_solution):
        """
        Initialise the particle with its initial position and velocity, and an instance of the pallet problem (used
        to generate and evaluate solutions). 

        Unlike in standard PSO, pbest is initialised to a position generated by a local search on the start position,
        and the velocity is initialised to the vector toward pbest.
        """
        self.pallet_problem = pallet_problem
        self.position = initial_solution

        self.pbest = self.pallet_problem.iterative_neighbourhood_search(self.position, 30)
        self.pbest_cost = self.pallet_problem.evaluate_cost(self.pbest)

        self.velocity = [None for x in range(len(self.position))]
        random_position = self.pbest

        # Set initial velocity to half the difference between a random position and the initial position
        for i in range(len(random_position)):
            self.velocity[i] = (random_position[i] - self.position[i]) / 2

    def calculate_new_velocity(self, gbest):
        """
        Calculate the new velocity vector for the particle.

        Parameters:
        gbest: the global best position found by the swarm

        Returns: the new velocity vector
        """
        new_velocity = [None for x in range(len(gbest))]
        
        for i in range(len(self.velocity)):
            new_velocity[i] = (self.INERTIAL_COEFFICIENT * self.velocity[i]) + self.COGNITIVE_COEFFICIENT * random.random() * (self.pbest[i] - self.position[i]) + self.SOCIAL_COEFFICIENT * random.random() * (gbest[i] - self.position[i])
            
        return new_velocity

    def calculate_new_position(self):
        """
        Use the particle's velocity to calculate its new position.

        Returns: the new position
        """
        new_position = [None for x in range(len(self.position))]

        for i in range(len(self.position)):
            new_position[i] = self.position[i] + self.velocity[i]

        return new_position

    def update_particle(self, gbest):
        """
        Update the particle's velocity and position and, if the new position is better, update pbest.

        Parameters:
        gbest: the global best position found by the swarm

        Returns: the new position of the particle
        """
        self.velocity = self.calculate_new_velocity(gbest)
        self.position = self.calculate_new_position()
        new_cost = self.pallet_problem.evaluate_cost(self.position)

        if new_cost < self.pbest_cost:
            self.pbest = self.position
            self.pbest_cost = new_cost

        return self.position