COAL_INDEX = 13 # coal
WASTE_INDEX = 36 # waste
import simworld
# Alias to save some typing
_t = simworld.translate
def PRINT_SEP():
print '-' * 80 # separator
#
# Load the world for searching
welt = simworld.World()
print 'Loading world...'
welt.load(savefile_path)
fabs_in_welt = [
welt.get_factory_at(i) for i in range(welt.get_factory_count())
]
assert len(fabs_in_welt) == welt.get_factory_count()
def is_producer_of_good(fab, indx=-1):
if indx == -1:
if filter_catg == -1: return len(fab.get_ausgang()) > 0
return filter_catg in [
g.get_typ().catg_index for g in fab.get_ausgang()
def simulation(self):
# initial genotypes (clonal but different for each group)
self.initial_population()
# for each generation: create world
for n in range(self.n_gen):
print("\ngeneration = {}/{}".format(n + 1, self.n_gen))
world = simworld.World(self.xmax, self.ymax, self.n_walls,
self.n_trees)
best_e = 0
best_group = None
# for each group (same world for all for each group in the same generation)
for ng in range(self.n_groups):
print("group: {}/{}".format(ng + 1, self.n_groups))
# create new agents and introduce them in the simworld
genotype = self.genotypes[ng]
world.allocate_agents(self.n_agents, genotype)
# simulate group lifetime
for tx in tqdm(range(self.lifetime)):
world.update()
# get and print results
agents_e = [ag.energy for ag in world.agents]
av_e = sum(agents_e) / len(world.agents)
print("agents energy: {}, average = {}".format(agents_e, av_e))
# save if best
if av_e > best_e:
best_e = av_e
best_group = world.agents
print("new best = {}".format(best_e))
# reset world for next group
world.reset()
# print best data for past generation
print("\nbest agents:")
for ag in best_group:
print("agent energy = {}".format(ag.energy))
print("average energy = {}".format(best_e))
# check and save if better than previous best (all generations)
if best_e > self.best_e:
self.best_cases.append([best_e, best_group, world])
self.best_e = best_e
print("best over all generations")
if self.anim_best:
simanimation.world_animation(self.lifetime, best_group,
world)
# animation for fixed steps
if self.anim_step:
if (n + 1) % self.anim_step == 0:
simanimation.world_animation(self.lifetime, best_group,
world)
# breeding
best_genotype = best_group[0].genotype
# copy the identic best genotype for one group
self.genotypes = [best_genotype]
for ng in range(self.n_groups - 1):
# mutation of the nnet only for now
# thresholds
ut = best_genotype.ut + np.random.randint(-1,
2) * self.mut_rate
lt = best_genotype.lt + np.random.randint(-1,
2) * self.mut_rate
vt = best_genotype.vt + np.random.randint(-1,
2) * self.mut_rate
# weights
W = self.mut_weights(best_genotype.W)
V = self.mut_weights(best_genotype.V)
# create and save new genotype
new_genotype = simgenotype.Genotype(ut=ut,
lt=lt,
vt=vt,
W=W,
V=V)
self.genotypes.append(new_genotype)