inteligenty-traktor/src/generate_field.py

117 lines
4.0 KiB
Python

from random import randint, choices, random
from kb import tractor_kb, multi_sasiedzi
import pytholog as pl
from numpy.random import choice as npchoice
def score_field(field):
score = 0
for index in range(len(field)):
neighbours = []
if index >= 16 and field[index-16] != 'water':
neighbours.append(field[index-16])
if index % 15 != 0 and field[index+1] != 'water':
neighbours.append(field[index+1])
if index < 240 and field[index+16] != 'water':
neighbours.append(field[index+16])
if index % 16 != 0 and field[index-1] != 'water':
neighbours.append(field[index-1])
mod = multi_sasiedzi(field[index], neighbours)[0]["Mul"]
if mod > 10:
print(mod, '= multi(', field[index], ', ', neighbours, ')')
score += mod
score = score / 256
return score
def choose_parents(population):
total_weights = sum(entity[0] for entity in population)
weights = [entity[0] / total_weights for entity in population]
selection = npchoice(len(population), size=2, replace=False, p=weights)
parents = [population[i] for i in selection]
return parents[0], parents[1]
def breed_and_mutate(mom, dad):
crossover_point = randint(1, len(mom[1]) - 2)
offspring = mom[1][:crossover_point] + dad[1][crossover_point:]
if len(offspring) != len(mom):
ValueError("offspring length is not equal to mom length")
if random() < 0.1:
mutation_index = randint(0, len(offspring) - 1)
while offspring[mutation_index] == 'water':
mutation_index = randint(0, len(offspring) - 1)
mutation = get_random_vegetable()
while mutation == offspring[mutation_index]:
mutation = get_random_vegetable()
offspring[mutation_index] = mutation
offspring_score = score_field(offspring)
# print('offspring score', offspring_score, 'for parents', mom[0], 'and', dad[0])
return [offspring_score, offspring]
def get_random_vegetable():
vegetables = [x['Nazwa_warzywa'] for x in tractor_kb.query(pl.Expr("warzywo(Nazwa_warzywa)"))]
return vegetables[randint(0,len(vegetables)-1)]
def genetic_algorithm(population, iterations):
population_size = len(population)
for entity in population:
entity[0] = score_field(entity[1])
for iteration in range(iterations):
population.sort(key=lambda x: x[0], reverse=True)
print('\n=====\n\n💪 Best individual in iteration', iteration, 'has a score of', population[0][0])
population = population[:population_size//2]
new_offspring = []
while len(population) + len(new_offspring) < population_size:
mom, dad = choose_parents(population)
child = breed_and_mutate(mom, dad)
new_offspring.append(child)
population.extend(new_offspring)
return population[0]
population = []
# each field has unmutable locations of water and grass tiles
water_tile_indexes = [1, 2, 3, 34, 37, 44, 45, 53, 60, 61, 69, 81, 82, 83, 84, 119, 120, 121, 136, 152, 187, 194, 202, 203, 204, 210, 219, 226, 227, 228]
grass_tile_indexes = [0, 39, 40, 56, 71, 72, 73, 86, 88, 114, 115, 130, 146, 147, 163, 164, 166, 167, 180, 181, 182, 231, 232, 233]
vegetables = [x['Nazwa_warzywa'] for x in tractor_kb.query(pl.Expr("warzywo(Nazwa_warzywa)"))]
for _ in range(100):
field = [vegetables[randint(0, 24)] for _ in range(256)]
for index in water_tile_indexes:
field[index] = "water"
for index in grass_tile_indexes:
field[index] = "grass"
# entities of the population are stored with two properties
# the first being the average score of the field
# and the second being the layout of the field
population.append([0, field])
best = genetic_algorithm(population, 20)
print('\n=====\n\nfinal field multiplier score is', best[0])
with open('field', 'w', encoding='utf-8') as file:
file.write(str(best[1]))
file.close
print('final field layout saved to file "field" in the current working directory\n')