def determine_fitness(self, population_size):
''' Determine Fitness
Returns a fitness value for each clone by determining performance in
simulation. Fitness is evaluated primarily based on race performance
(time, distance), but undesireable behaviors such as letting the battery
voltage stay too low will be punished with a decrease in fitness value.
'''
self.environment = para.environment # **temporary replacement for environment**
self.performances = []
self.distances = []
# runs each clone through simulation to determine its fitness
for n in range(population_size):
# export clone's network parameters and environment to set up a race
competition = race.Race(self.clones[n], self.environment)
# runs through a race allowing clone network to determine strategy
competition.race()
# aborts evolution proccess if a fatal error occurs
self.abort_evolution = competition.abort
# evaluate performance
performance = competition.argo.position**2 / competition.argo.race_time / para.environment[
0]
self.performances.append(performance)
print('average performance: ',
np.sum(self.performances) / population_size)
def main():
# STEP 1:
print('Starting Race')
data = scrape_data.log_data(1, progress_keys())
race_stats = race.Race(len(data))
musher_objects = []
for musher in data:
musher_objects.append(mushers.Mushers(musher))
update_insert_sql.start_race(musher_objects)
# STEPS 2-6:
# list of the logs we'll be using in the sim
log_nums = sim_log_nums()
print('Working log #1: 1 of %s' % (len(log_nums)+1))
sleep(5*60)
for log in log_nums[1:]:
print('Working log #%s: %s of %s' % (log, (log_nums.index(log) +1), (len(log_nums) +1)))
# STEP 2:
updated_list = scrape_data.log_data(log, progress_keys())
musher_list = []
# STEP #3:
for musher in updated_list:
new_musher = mushers.Mushers(musher)
race_stats.add_checkpoints(musher['Checkpoint'])
new_musher.tally_points(musher, race_stats.get_num_mushers(), race_stats.get_dog_checks(), race_stats.reached_Nome())
musher_list.append(new_musher)
# STEP 4:
update_insert_sql.update_race(musher_list)
# STEP 5:
sleep(5*60)
print('Race Finished')
import race
import turn
import ship
import fabrication
# gal = galaxy.Galaxy()
galaxy.gal.init_gen()
game_exit = 0
while game_exit != 1 :
galaxy.gal.galaxy2html("graph")
action = input("input command: ")
if action == "makeciv":
race.Race(galaxy.gal)
elif action == "turn" or action == "t":
turn.gala_time.time_turn()
elif action == "cheatship":
has_ai = input("0 if no ai, anything else if it is: ")
sys = None
for system in galaxy.gal.systems:
if system.planets:
for planet in system.planets:
if planet.owner != None:
sys = system
break
if sys is None:
continue
new_ship = ship.Ship(sys)
if not has_ai:
def main():
'''Assumes the Iditarod has started, and updates the table'''
# get the available logs
original_logs = get_all_logs()
# if logs are empty, try again in 5
while not original_logs:
original_logs = get_all_logs()
sleep(5 * 60)
# ----------------------------------------------------------------------------------------
# RACE HAS BEGUN BABY
# ----------------------------------------------------------------------------------------
# give a time out to the while loop
timeout = 24 * 60 * 60
data = scrape_data.log_data(1, progress_keys())
race_stats = race.Race(len(data))
musher_objects = []
for musher in data:
musher_objects.append(mushers.Mushers(musher))
update_insert_sql.start_race(musher_objects)
# STEPS 2-6:
# list of the logs we'll be using in the sim
new_logs = get_all_logs()
print('Working log #1')
sleep(5 * 60)
lindex = 0
while original_logs != new_logs or timeout:
# log
log = new_logs[lindex]
if original_logs == new_logs:
timeout -= 5 * 60
sleep(5 * 60)
# skip already used logs
elif log in original_logs:
lindex += 1
continue
else:
# reset timeout
timeout = 24 * 60 * 60
print('Working log #%s' % log)
# STEP 2:
updated_list = scrape_data.log_data(log, progress_keys())
race_stats.add_checkpoints(updated_list[0]['Checkpoint'])
musher_list = []
# STEP #3:
for musher in updated_list:
new_musher = mushers.Mushers(musher)
new_musher.tally_points(musher, race_stats.get_num_mushers(),
race_stats.get_dog_checks(),
race_stats.reached_Nome())
musher_list.append(new_musher)
# STEP 4:
update_insert_sql.update_race(musher_list)
# STEP 5:
sleep(5 * 60)
# add used log to original
original_logs.append(log)
# add any new logs to the new_log list
added_logs = get_all_logs()
for l in added_logs:
if l not in new_logs:
new_logs.append(l)
lindex += 1
# wait three weeks before deleting the table
sleep(60 * 60 * 24 * 7 * 3)
update_insert_sql.clear_table()
async def restart(ctx):
if ctx.channel.id in races:
await ctx.send(msg.MSG_RE_START)
else:
await ctx.send(msg.MSG_RACE_CREATE)
races[ctx.channel.id] = race.Race(ctx.channel.id)
async def start(ctx):
if ctx.channel.id not in races:
await ctx.send(msg.MSG_RACE_CREATE)
races[ctx.channel.id] = race.Race(ctx.channel.id)
else:
await ctx.send(msg.MSG_CONF_EXIST)
def reproduce(self, selection_bias, inheritance_rate, population_size,
generation):
''' Reproduce - creates a child network based on clone performance
A new mutation equal to the weighted average of ckone mutations
is applied to the original network. Weights are proportional
to (clone performance[i] - average clone performance)^selection_bias
Inheritance rate scales how big the new mutation is.
'''
# Create list of weights
mean_performance = np.sum(self.performances) / population_size
weights = [(((performance - mean_performance)) / mean_performance)**
selection_bias * inheritance_rate
for performance in self.performances]
# Initialize child network
child_net = self.parameters[:]
# Loop through the numpy arrays that make up the parameters
for parameter_type in [0, 1]: # No 2 index to preserve self.sizes
for index in range(len(self.sizes) - 1):
# Initialize np array from 1st clone **required to sum**
parameter_deltas = (self.mutations[0][parameter_type][index] *
weights[0] * inheritance_rate)
# Add weighted np arrays from rest of clones,
for clone in range(1, population_size):
parameter_deltas = (
parameter_deltas +
self.mutations[clone][parameter_type][index] *
weights[clone] * inheritance_rate)
# Normalize by population size and update child_net
parameter_deltas = parameter_deltas / float(population_size)
child_net[parameter_type][index] = (child_net[parameter_type]\
[index] + parameter_deltas)
# Determine and display performance of child network
competition = race.Race(child_net, self.environment)
if generation % 200 == True:
competition.race(True)
print("show graph")
plt.figure(1, figsize=(30, 50))
plt.subplot(611)
np_battery_tracker = np.array(
competition.battery_tracker) / para.battery_max_charge
plt.title("battery")
plt.plot(np_battery_tracker)
plt.subplot(612)
plt.title("distance")
plt.plot(competition.distance_tracker)
plt.subplot(613)
plt.title("velocity")
plt.plot(competition.velocity_tracker)
plt.subplot(614)
plt.title("irradiance")
plt.plot(competition.irradiance_tracker)
plt.subplot(615)
#plt.title("rolling loss")
#plt.plot(competition.rolling_loss_tracker)
#plt.subplot(616)
#plt.title("aero loss")
#plt.plot(competition.aero_loss_tracker)
plt.savefig("graphs.png")
else:
competition.race()
print(
'new gen: ', competition.argo.position**2 /
competition.argo.race_time / para.environment[0])
vector.sub(vector.sub(dir, stop), self.velocity))
return vector.normalize(vector.sub(dir, stop))
class TweakedAI(race.BaseAI):
def __init__(self, scale: float = 1.0):
super().__init__()
self.color = (40 * scale, 0, 255)
self.scale = scale
def evaluate(self, goal, goal2):
dir = vector.normalize(vector.from_to(self.position, goal))
stop = vector.normalize(self.velocity)
stop = vector.mult(stop, vector.dot(stop, dir) * self.scale)
return vector.normalize(vector.sub(dir, stop))
def __repr__(self):
return "TweakedAI (%.1f)" % self.scale
if __name__ == "__main__":
r = race.Race(time=40)
r.add_racer(EasyAI())
r.add_racer(DragAI(1))
r.add_racer(DragAI(2))
r.add_racer(PerpAI())
r.add_racer(CombinedAI())
r.add_racer(StoppingAI())
r.add_racer(TweakedAI(1))
r.start_pygame_race()