def move_to_food(self, foods):
"""
Selection of Food to move towards, and ant movement
:param foods: (list) Food objects in noticeable objects
:return: (array) new ant position
"""
# Go directly to food if there is only one source
if len(foods) == 1:
return self.move_to(foods[0].position)
# Compare food sources in terms of size and distance to nest
else:
sub_food = []
for i, obj in enumerate(foods):
# If food size equal to zero, we ignore the Food object
if obj.size == 0:
pass
# If distance is equal to zero, we ignore the Food object
elif distance(obj.position - self.home.position) == 0:
pass
else:
sub_food.append(obj)
if sub_food:
# Getting features of food objects
data = zeros((len(sub_food), 2))
for i, obj in enumerate(sub_food):
# Food size
data[i, 0] = obj.size
# Distance to nest
data[i, 1] = distance(obj.position - self.home.position)
# Rescaling each feature to have values bounded by 1
data /= np.max(data, axis=0)
# Calculating probability distribution
probs = (data[:, 0] ** self.foodiness) * (data[:, 1] ** self.explorativeness)
probs /= np.sum(probs)
# Drawing an object from the prob distribution
index = np.random.choice(len(sub_food), p=probs)
# index = np.argmax(probs)
return self.move_to(sub_food[index].position)
else:
return None
def move_to_pheromone(self, pheromones):
"""
Selection of Pheromone to move towards, and ant movement
:param pheromones: (list) Pheromone objects in noticeable objects
:return: (array) new ant position
"""
# Go directly to scent if there is only one source
if len(pheromones) == 1:
return self.move_to(pheromones[0].position)
# Compare pheromone sources
else:
# Getting features of pheromone objects
data = zeros((len(pheromones), 3))
for i, obj in enumerate(pheromones):
# Intensity
data[i, 0] = obj.strength
# Distance to nest
data[i, 1] = distance(obj.position - self.home.position)
# Difference in momentum
# TODO define difference in momentum
data[i, 2] = 1
# Rescaling each feature to have values bounded by 1
data /= np.max(data, axis=0)
# Calculating probability distribution
probs = (data[:, 0] ** self.inscentiveness) * (data[:, 1] ** self.explorativeness)
probs *= (data[:, 2] ** self.directionism)
probs /= np.sum(probs)
# Draw an object from the prob distribution
index = np.random.choice(len(pheromones), p=probs)
# index = np.argmax(probs)
return self.move_to(pheromones[index].position)
def reset_parameters(self):
glorot(self.weight)
zeros(self.bias)