def StartEvolution(
self,
name,
date,
type,
description,
flag,
num_population,
num_robots,
num_levels,
num_gen,
block_size,
gen_threshold,
actual_gen,
player,
mutation,
cross,
wrld,
):
# ATENCAO!! Deve existir pelo menos 1 GameWorld
# depois arrumo essa verificacao no script do banco, mas agora PRECISA ser assim
# get World config
gameworld = GameWorld.objects.get(id=wrld)
my_world = World(gameworld.m, gameworld.n)
my_world.MaxArea(gameworld.max_areas)
my_world.MaxUnits(gameworld.max_units)
my_world.Production(gameworld.prod_unit0, gameworld.prod_unit1)
my_world.Costs(gameworld.cost_gateway, gameworld.cost_unit0, gameworld.cost_unit1)
for area in Area.objects.all().filter(world=gameworld.id):
my_world.CreateArea(area.x, area.y, area.length)
# declare Genetic Algorithm for the problem
ga = GA(my_world, cross, mutation, num_gen, num_population)
pareto = tools.ParetoFront()
pop = []
# create a new exp
now = datetime.datetime.now()
exp = Experiment(
world=gameworld,
name=name,
date=date,
block_size=block_size,
start=now,
type=type,
description=description,
flag=flag,
actual_gen=actual_gen,
CXPB=cross,
MUTPB=mutation,
NGEN=num_gen,
NPOP=num_population,
num_robots=num_robots,
numLevels=num_levels,
gen_threshold=gen_threshold,
numMinPlayers=player,
time_elapsed_end=0,
)
exp.save()
# set NEW Population
pop = ga.SetPopulation()
# set block for the generation block table
num_block = 1
# set FITNESS to the entire population
fitnesses = ga.GetFitness(pop)
ga.AttachFitness(pop, fitnesses)
# assign the crowding distance to the individuals
ga.SetCrowdingDistance(pop)
# discover again the LAST Pareto
pareto.update(pop)
# loop for BLOCK generations
for i in range(1, exp.block_size):
# Select the next generation individuals
# ga.Selection(pop, my_world)
ga.NSGASelection(pop, my_world)
pareto.update(pop)
# update actual generations with block_size
exp.actual_gen += exp.block_size
# save Generations Block
gen = Generation(experiment=exp, block=num_block, comparisons="to be collected")
gen.save()
# save population and Front into database
with transaction.atomic():
count = 1
for ind in pop:
population = Population(generation=gen, chromosome=str(ind[0]), index=count)
population.save()
count += 1
count = 1
for ind in pareto:
pfront = PFront(generation=gen, chromosome=str(ind[0]), index=count)
pfront.save()
count += 1
# set the next status according the number of generations
if exp.actual_gen >= exp.NGEN:
exp.flag = "F"
else:
exp.flag = "R"
exp.save()
# run robots
self.RunRobots(exp, num_block, ga, my_world)
return exp
def ContinueEvolution(self):
# get last ID of the Experiment
exp = Experiment.objects.latest("id")
# ATENCAO!! Deve existir pelo menos 1 GameWorld
# depois arrumo essa verificacao no script do banco, mas agora PRECISA ser assim
# inclusive fixei HARDCODE que sera esse mundo apenas
# get World config
gameworld = GameWorld.objects.get(id=1) ###HARDCODE
my_world = World(gameworld.m, gameworld.n)
my_world.MaxArea(gameworld.max_areas)
my_world.MaxUnits(gameworld.max_units)
my_world.Production(gameworld.prod_unit0, gameworld.prod_unit1)
my_world.Costs(gameworld.cost_gateway, gameworld.cost_unit0, gameworld.cost_unit1)
for area in Area.objects.all().filter(world=gameworld.id):
my_world.CreateArea(area.x, area.y, area.length)
# declare Genetic Algorithm for the problem
ga = GA(my_world, exp.CXPB, exp.MUTPB, exp.NGEN, exp.NPOP)
pareto = tools.ParetoFront()
pop = []
# check if it is START or CONTINUE evolution - must know if I will create a new experiment or use the one received
# if F then start NEW
# if exp.flag=='F':
#
# #create a new exp
# now=datetime.datetime.now()
# exp = Experiment(world=gameworld, name='Bangkok experimento ' + now.strftime("%Y-%m-%d %H:%M:%S"), date=now, block_size=10,
# flag='W', actual_gen=0, keep_interaction=True, CXPB=0.3, MUTPB=0.1, NGEN=200, NPOP=300)
# exp.save()
#
#
# #set NEW Population
# pop=ga.SetPopulation()
#
#
# #set block for the generation block table
# num_block=1
#
# else:
# number of the next block of generations
gen = Generation.objects.latest("id")
num_block = gen.block + 1
# fake population
pop = ga.SetPopulationFake(my_world)
# get the LAST Population
count = 0
for ind in Population.objects.all().filter(generation=gen.id):
pop[count][0] = literal_eval(ind.chromosome)
count += 1
###### EM ########
# set FITNESS to the entire population
fitnesses = ga.GetFitness(pop)
ga.AttachFitness(pop, fitnesses)
# assign the crowding distance to the individuals
ga.SetCrowdingDistance(pop)
# discover again the LAST Pareto
pareto.update(pop)
#
if exp.actual_gen + exp.block_size <= exp.gen_threshold and num_block <= exp.numLevels:
loop = exp.block_size
else:
loop = exp.NGEN - exp.actual_gen
# loop for BLOCK generations
for i in range(1, loop):
# Select the next generation individuals
# ga.Selection(pop, my_world)
ga.NSGASelection(pop, my_world)
pareto.update(pop)
# update actual generations with block_size
exp.actual_gen += loop
# save Generations Block
gen = Generation(experiment=exp, block=num_block, comparisons="to be collected")
gen.save()
# save population and Front into database
with transaction.atomic():
count = 1
for ind in pop:
population = Population(generation=gen, chromosome=str(ind[0]), index=count)
population.save()
count += 1
count = 1
for ind in pareto:
pfront = PFront(generation=gen, chromosome=str(ind[0]), index=count)
pfront.save()
count += 1
# set the next status according the number of generations
if exp.actual_gen >= exp.NGEN:
exp.flag = "F"
else:
exp.flag = "R"
exp.save()
# run robots
self.RunRobots(exp, num_block, ga, my_world)