def test_mutate():
"""
Mutate 5 indiviuals with identical sequences and make sure
that they are mutated in a random way.
"""
sequence = "0123456789ABCDEF"
individual1 = Chromosome(sequence=sequence)
individual2 = Chromosome(sequence=sequence)
individual3 = Chromosome(sequence=sequence)
individual4 = Chromosome(sequence=sequence)
individual5 = Chromosome(sequence=sequence)
mutated1 = mutate(individual1, 1.0)
mutated2 = mutate(individual2, 1.0)
mutated3 = mutate(individual3, 1.0)
mutated4 = mutate(individual4, 1.0)
mutated5 = mutate(individual5, 1.0)
assert mutated1.sequence != individual1.sequence
assert mutated2.sequence != individual2.sequence
assert mutated3.sequence != individual3.sequence
assert mutated4.sequence != individual4.sequence
assert mutated5.sequence != individual5.sequence
assert mutated1.sequence != individual2.sequence
assert mutated1.sequence != individual3.sequence
assert mutated1.sequence != individual4.sequence
assert mutated1.sequence != individual5.sequence
def test_no_crossover():
parent1 = Chromosome(sequence="0123456789ABCDEF")
parent2 = Chromosome(sequence="FEDCBA9876543210")
offspring1, offspring2 = crossover(parent1, parent2, 0.0)
assert offspring1.sequence == parent1.sequence
assert offspring1.sequence != parent2.sequence
assert offspring2.sequence != parent1.sequence
assert offspring2.sequence == parent2.sequence
assert offspring1.sequence != offspring2.sequence
def test_different_length_crossover():
individual1 = Chromosome(sequence=get_random_sequence(100))
individual2 = Chromosome(sequence=get_random_sequence(10))
offspring1, offspring2 = crossover(individual1, individual2, 1.0)
assert offspring1.sequence != individual1.sequence
assert offspring1.sequence != individual2.sequence
assert offspring2.sequence != individual1.sequence
assert offspring2.sequence != individual2.sequence
print(individual1.sequence)
print(offspring1.sequence)
print(individual2.sequence)
print(offspring2.sequence)
def main():
"""
Run many iterations and evolve the best combination of keys and time
"""
parser = argparse.ArgumentParser(description='''
Run a genetic algorithm to play the game qwop''')
parser.add_argument('--save', '-s', action='store_true')
parser.add_argument('--seed')
args = parser.parse_args()
pop_size = 20
population = []
max_play_time = 20
crossover_rate = 0.7
mutation_rate = 0.05
generations = 0
fittest = 0.0
total_fitness = 0.0
max_chromo_length = 40
fittest_ind = 0
goal = int(max_play_time * 0.8)
image_counter = 0
step_execute_time = 0.1
run_instance = datetime.datetime.now().isoformat()
try:
if args.seed:
for _ in range(pop_size):
#Create initial population
seed_chromo = Chromosome(sequence=args.seed, fitness=0.0)
mutated_seed = mutate(seed_chromo, mutation_rate)
population.append(
Chromosome(sequence=mutated_seed.sequence, fitness=0.0))
else:
for _ in range(pop_size):
#Create initial population
population.append(
Chromosome(sequence=get_random_sequence(
random.randrange(max_chromo_length,
max_chromo_length + 1)),
fitness=0.0))
while fittest < goal:
auto_qwop = AUTOQWOP()
auto_qwop.load_website()
auto_qwop.get_game()
total_fitness = 0.0
fittest = 0.0
for i, chromo in enumerate(population):
#Loop over each member of the population
game_over = False
game_won = False
timed_out = False
start_time = time.time()
iters = 0
current_frame = None
time.sleep(1)
while not game_over and not game_won and not timed_out:
#Keep looping until game has ended
for state in chromo.sequence:
step_start_time = time.time()
#Loop over the steps in the sequence
auto_qwop.update_outputs(state)
if iters == 0 or iters % 30 == 0:
current_frame = auto_qwop.get_frame()
game_over = auto_qwop.test_for_game_over(
current_frame)
game_won = auto_qwop.test_for_game_won(
current_frame)
if args.save:
current_frame.save(
'./play_images/{}.png'.format(
image_counter), 'PNG')
image_counter += 1
if time.time() - start_time >= max_play_time:
timed_out = True
if game_over or game_won or timed_out:
break
iters += 1
execution_time = time.time() - step_start_time
if execution_time < step_execute_time:
time.sleep(step_execute_time - execution_time)
#print(time.time() - step_start_time)
auto_qwop.update_outputs('0')
current_frame = auto_qwop.get_frame()
run_time = time.time() - start_time
distance = auto_qwop.get_distance(current_frame)
speed = distance / run_time
if game_won:
population[i].fitness = distance
if game_over or timed_out:
population[i].fitness = distance
if population[i].fitness > fittest:
fittest = population[i].fitness
fittest_ind = i
# print('Ran {} meters in {} seconds!'.format(distance,
# run_time))
total_fitness += population[i].fitness
if timed_out:
auto_qwop.update_outputs('2')
time.sleep(0.5)
auto_qwop.update_outputs('4')
time.sleep(0.5)
auto_qwop.update_outputs('1')
#Commit "suicide" so we don't have to refresh the page
#every time. Saves time and bandwidth.
suicide_attempts = 0
while not game_over and suicide_attempts < 10:
game_over = auto_qwop.test_for_game_over(
auto_qwop.get_frame())
suicide_attempts += 1
time.sleep(1)
auto_qwop.update_outputs('0')
if game_over:
auto_qwop.restart()
else:
auto_qwop.driver.refresh()
auto_qwop.get_game()
else:
auto_qwop.restart()
auto_qwop.driver.quit()
temp_pop = []
with open('{}.txt'.format(run_instance), 'a') as f:
print('Generation {}, {}, {}, {}'.format(
generations, total_fitness,
population[fittest_ind].sequence,
population[fittest_ind].fitness),
file=f)
while len(temp_pop) < pop_size:
"""
Select offspring from the population via roulette wheel
selection and create a new population after mating
and mutating them.
"""
parent1 = roulette(total_fitness, population)
parent2 = roulette(total_fitness, population)
offspring1, offspring2 = crossover(parent1, parent2,
crossover_rate)
offspring1 = mutate(offspring1, mutation_rate)
offspring2 = mutate(offspring2, mutation_rate)
offspring1.fitness = 0.0
offspring2.fitness = 0.0
temp_pop.append(offspring1)
temp_pop.append(offspring2)
#copy temp_pop to population before starting all over again
population = temp_pop
print("Generation {}".format(generations))
print("Total fitness = {}".format(total_fitness))
generations += 1
print("You did it!")
with open('sequence.txt', 'a') as f:
print(population[fittest_ind].sequence, file=f)
print(population[fittest_ind].fitness, file=f)
except Exception as e:
print(e)
raise (e)
finally:
#clean up
auto_qwop.driver.quit()
def test_no_mutate():
sequence = "123456789ABCDEF"
individual = Chromosome(sequence=sequence)
mutated = mutate(individual, 0.0)
assert mutated.sequence == individual.sequence
def test_chromo():
sequence = "2355625130"
fitness = 1.2
chromo = Chromosome(sequence=sequence, fitness=fitness)
assert chromo.sequence == sequence
assert chromo.fitness == fitness