def search(city):
visited = []
best_score = None
best_move = None
for i in range(iterations):
r = random()
if r <= density_change_chance:
idx = dens = -1
lmt = 0 #RZ limit the while loop, it will cause dead loop when density_change_chance is high
while ((dens == -1 or idx == -1) \
or ("DENSITY", idx, dens) in visited) \
and lmt < 6 * 30 : #RZ possible moves
idx = randint(id_range[0],
id_range[1] - 1) #TOTO magic number here?
dens = randint(density_range[0], density_range[1])
lmt = lmt + 1 #RZ
mov = ("DENSITY", idx, dens)
else:
x = y = newid = -1
lmt = 0 #RZ limit the while loop
while ((x == -1 or y == -1 or newid == -1) \
or ("CELL", x, y, newid) in visited) \
and lmt < 256 * 6 : #RZ possible moves
x = randint(0, city.width - 1)
y = randint(0, city.height - 1)
newid = randint(id_range[0], id_range[1])
lmt = lmt + 1 #RZ
mov = ("CELL", x, y, newid)
visited.append(mov)
scr = score(city, mov)
if best_score == None or scr > best_score:
best_score = scr
best_move = mov
suggested_city = move(city, best_move)
# Update AI params based on this move - changes from Ryan
log.info("AI search complete. Best score = {}. Best move = {}.".format(
best_score, best_move))
suggested_city.updateAIMov(best_move)
return (suggested_city, best_move, objective.get_metrics(suggested_city))
def metrics(self):
return metrics_dictionary(objective.get_metrics(self))
def search(city, queue=None):
"""Random single moves AI search algorithm.
Args:
city (cityiograph.City): the city for which we would like to optimize our metrics
visited (set, optional): describes previously visited states in our prediction
Returns:
3-tuple: suggested_city (cityiograph.City): -
best_move (3-tuple): move type { 'DENSITY', 'CELL' } (str)
index (int) and new density (int)
OR
x (int), y (int) and new type id for that cell
scores (list): AI metrics scores
"""
if queue is None:
queue = set()
visited = copy.copy(queue)
best_score = None
best_scores = None # RZ 170615 passing score array to json
best_move = None
# RZ 170615 update the weights before search
objective.update_weights(city.AIWeights)
for i in range(iterations):
r = random.random()
if r <= density_change_chance:
idx = dens = -1
lmt = 0 # RZ limit the while loop, it will cause dead loop when density_change_chance is high
while ((dens == -1 or idx == -1)
or ("DENSITY", idx) in visited) \
and lmt < 6 * 30 * 4: # RZ possible moves * 4
# TOTO magic number here?
idx = random.randint(id_range[0], id_range[1])
dens = random.randint(density_range[0], density_range[1])
lmt = lmt + 1 # RZ
mov = ("DENSITY", idx, dens)
else:
x = y = newid = -1
lmt = 0 # RZ limit the while loop
while ((x == -1 or y == -1 or newid == -1)
or x < 4 or x > 12 or y < 4 or y > 12 or x == 8 or y == 8
or ("CELL", x, y, newid) in visited) \
and lmt < 256 * 6 * 4: # RZ possible moves * 4 # Focus center of city and no road cell
x = random.randint(0, city.width - 1)
y = random.randint(0, city.height - 1)
newid = random.randint(id_range[0], id_range[1])
lmt = lmt + 1 # RZ
mov = ("CELL", x, y, newid)
# Add this move to the visited set
if mov[0] == 'DENSITY':
visited.add(mov[:-1])
else:
visited.add(mov)
# KL - minor optimization to avoid duplicate calls to score method
[scr, best] = scores(city, mov)
if best_score is None or scr > best_score:
best_score = scr
best_scores = best
best_move = mov
# Determine our suggested city based on this move
suggested_city = move(city, best_move)
# Run the ML prediction on this suggested city
final_city = ML.predict(suggested_city)
# Update AI params based on this move - changes from Ryan
log.info("AI search complete. Best score = {}. Best move = {}.".format(
best_score, best_move))
final_city.AIMov = best_move
final_city.score = scr
return final_city, best_move, objective.get_metrics(final_city)
unity_server.send_data(result)
log.info(
"Same city received, but new toggle value. New ml_city and ai_city data successfully sent to GH.\n")
log.info("Waiting to receive new city...")
else:
# RZ 170626 same city received and not the first city, update
# meta data for GH UI, do not write to local file
# RZ 170614 update city.animBlink
ml_city.animBlink = animBlink
ai_city.animBlink = animBlink
# Then, get metrics dicts for cities
ml_metrics = metrics_dictionary(objective.get_metrics(ml_city))
ai_metrics = metrics_dictionary(ai_metrics_list)
# RZ firstly, we need to update only the meta data of the 2
# cities, including slider position and AI Step
ml_city.updateMeta(input_city)
ai_city.updateMeta(input_city)
# Send resulting 2-city dictionary (predict/ai) back to GH
if input_city.AIStep < 20: # KL - accounting for AI step case
result = {'predict': ml_city.to_dict(), 'ai': None}
else:
result = {'predict': ml_city.to_dict(),
'ai': ai_city.to_dict()}
server.send_data(result)
unity_server.send_data(result)