Traktor/source/genetic.py

99 lines
3.2 KiB
Python
Raw Permalink Normal View History

import random
def make_population(population_s, field_s):
population = []
crops = ['apple', 'cauliflower', 'radish', 'wheat', 'rock_dirt', 'dirt']
for _ in range(population_s):
i = []
for _ in range(field_s):
row = random.choices(crops, k=field_s)
i.append(row)
population.append(i)
return population
def calculate_fitness(individual):
cost = 0
for i in range(len(individual)):
for j in range(len(individual[i])):
crop = individual[i][j]
neighbors = [
individual[x][y]
for x in range(max(0, i-1), min(len(individual), i+2))
for y in range(max(0, j-1), min(len(individual), j+2))
if (x,y) != (i,j)
]
for n in neighbors:
if crop == 'wheat' and n == 'apple':
cost += 2
elif crop == 'cauliflower' and n == 'radish':
cost += 4
fitness = 1/(1+cost)
return fitness
def select_parents(population, fitnesses):
fitnesses_sum = sum(fitnesses)
selection_parts = [fitness / fitnesses_sum for fitness in fitnesses]
parents = random.choices(population, weights=selection_parts, k=2)
return parents
def crossover(parent_1, parent_2):
crossover_point = random.randint(1, (len(parent_1)-1))
child_1 = parent_1[:crossover_point] + parent_2[crossover_point:]
child_2 = parent_2[:crossover_point] + parent_1[crossover_point:]
return child_1, child_2
def mutation(individual, chance):
crops = ['apple', 'cauliflower', 'radish', 'wheat', 'rock_dirt', 'dirt']
if random.random() < chance:
row = random.randint(0, len(individual) - 1)
column = random.randint(0, len(individual[0]) - 1)
individual[row][column] = random.choice(crops)
return individual
def genetic_algorithm(population_s, field_s, chance, limit):
population = make_population(population_s, field_s)
best_fitness = 0
count = 0
while best_fitness < 1:
fitnesses = [calculate_fitness(individual) for individual in population]
new_population = []
for _ in range(population_s // 2):
parent_1, parent_2 = select_parents(population, fitnesses)
p1c = calculate_fitness(parent_1)
p2c = calculate_fitness(parent_2)
print("p1c: ",p1c,"\np2c: ",p2c)
child_1, child_2 = crossover(parent_1, parent_2)
child_1 = mutation(child_1, chance)
child_2 = mutation(child_2, chance)
new_population.append(child_1)
new_population.append(child_2)
combined_population = population + new_population
combined_population = sorted(combined_population, key=calculate_fitness, reverse=True)
population = combined_population[:population_s]
current_best_fitness = calculate_fitness(population[0])
if current_best_fitness > best_fitness:
best_fitness = current_best_fitness
count = 0
else:
count += 1
if count >= limit:
break
best_child = max(population, key=calculate_fitness)
bsf = calculate_fitness(best_child)
print("bsf: ", bsf)
return best_child