parallelHillClimber.py

import pybullet as p
from solution import SOLUTION
import constants as c
import copy
import os
import pickle

class PARALLEL_HILL_CLIMBER:

    def __init__(self):
        os.system("rm brain*.nndf")
        os.system("rm fitness*.txt")
        os.system("rm body*.urdf")

        self.parents = {}
        self.nextAvailableID = 0
        for i in range(0, c.populationSize):
            self.parents[i] = SOLUTION(self.nextAvailableID)
            self.nextAvailableID += 1
        # self.parent = SOLUTION()
        # print("Out HILL_CLIMBER")
        self.bestFitnessValues = []

    def Evolve(self, folderPath):
        self.Evaluate(self.parents)
        # self.parent.Evaluate("GUI")
        for currentGeneration in range(0, c.numberOfGenerations):
            self.Evolve_For_One_Generation(currentGeneration)
            if currentGeneration % c.gensPerPickle == 0:
                lowestFitnessParent = self.Get_Best_Robot()
                pickle.dump( lowestFitnessParent.linkPlan, open( folderPath + "Gen" + str(currentGeneration) + "linkPlan.p", "wb" ) )
                pickle.dump( lowestFitnessParent.jointPlan, open( folderPath + "Gen" + str(currentGeneration) + "jointPlan.p", "wb" ) )
                pickle.dump( lowestFitnessParent.sensorWeights, open( folderPath + "Gen" + str(currentGeneration) + "sensorWeights.p", "wb" ) )
                pickle.dump( lowestFitnessParent.motorWeights, open( folderPath + "Gen" + str(currentGeneration) + "motorWeights.p", "wb" ) )
                pickle.dump( lowestFitnessParent.numHiddenNeurons, open( folderPath + "Gen" + str(currentGeneration) + "hiddenNeurons.p", "wb" ) )

    def Evolve_For_One_Generation(self, generation):
        self.Spawn()
        self.Mutate(generation)
        self.Evaluate(self.children)
        # self.Print()
        self.Select()

    def Spawn(self):
        self.children = {}
        for id in self.parents:
            self.children[id] = copy.deepcopy(self.parents[id])
            self.children[id].Set_ID(self.nextAvailableID)
            self.nextAvailableID += 1

    def Mutate(self, generation):
        for id in self.children:
            self.children[id].Mutate(generation)

    def Select(self):
        minFitness = 10000
        for id in self.parents:
            if self.children[id].fitness < self.parents[id].fitness:
                self.parents[id] = self.children[id]

            if self.parents[id].fitness < minFitness:
                minFitness = self.parents[id].fitness

        self.bestFitnessValues.append(minFitness)

    def Print(self):
        for id in self.parents:
            print("")
            print('Parent fitness: ' + str(self.parents[id].fitness))
            print('Child fitness: ' + str(self.children[id].fitness))
            print("")

    def Get_Best_Robot(self):
        lowestFitnessParent = self.parents[0]
        for id in self.parents:
            if self.parents[id].fitness < lowestFitnessParent.fitness:
                lowestFitnessParent = self.parents[id]

        return lowestFitnessParent

    def Show_Best(self):
        lowestFitnessParent = self.Get_Best_Robot()
        lowestFitnessParent.Start_Simulation("GUI")
        lowestFitnessParent.Wait_For_Simulation_To_End()

    def Evaluate(self, solutions):
        for id in solutions:
            solutions[id].Start_Simulation("DIRECT")
        for id in solutions:
            solutions[id].Wait_For_Simulation_To_End()