def agglomerative(ex):
"""
@param ex: Explorer
"""
row_names = []
col_names = []
people_with_cars = [i for i, x in enumerate(ex.people_capacity) if x > 0]
people_without_cars = [
i for i, x in enumerate(ex.people_capacity) if x == 0
]
# initiate matrix with non-car people on rows and car people on columns
first_i = -1
distance_array = []
assignments_dict = {}
# initialize dictionary of assignments_dict
#car_centroids = {}
for car_person in people_with_cars:
assignments_dict[car_person] = []
#car_centroids[car_person] = ex.locations[car_person]
#initialize matrix of distances between non-car people to car people
for i in people_with_cars:
person_dist = []
row_names.append(i)
if first_i == -1:
first_i = i
for j in people_without_cars:
if (i == first_i):
col_names.append(j)
person_dist.append(ex.distances[i][j]['weight'])
distance_array.append(person_dist)
distance_matrix = np.matrix(distance_array)
if ex.verbose:
print "distance_matrix"
print distance_matrix
# while there is an unassigned non-car person, assign the person with minimum distance to a car with space. Then recenter the location of that car.
while len(col_names) > 0:
min_row = np.unravel_index(np.argmin(distance_matrix),
np.shape(distance_matrix))[0]
min_col = np.unravel_index(np.argmin(distance_matrix),
np.shape(distance_matrix))[1]
min_car = row_names[min_row]
assignments_dict[min_car].append(col_names[min_col])
distance_matrix = np.delete(distance_matrix, min_col, 1)
del col_names[min_col]
if len(assignments_dict[min_car]) >= (ex.people_capacity[min_car] - 1):
distance_matrix = np.delete(distance_matrix, min_row, 0)
del row_names[min_row]
#change the location of the car to be centroid of all people assigned to it and recompute distances for that row
else:
centroid_x, centroid_y = ex.locations[min_car]
for person in assignments_dict[min_car]:
centroid_x += ex.locations[person][0]
centroid_y += ex.locations[person][1]
centroid_x = centroid_x / (len(assignments_dict[min_car]) + 1)
centroid_y = centroid_y / (len(assignments_dict[min_car]) + 1)
new_row_dist = []
for person in col_names:
new_row_dist.append(
((ex.locations[person][0] - centroid_x)**2 +
(ex.locations[person][1] - centroid_y)**2)**0.5)
distance_matrix[min_row] = new_row_dist
total_cost = 0.0
assignment = [None] * ex.num_people
for driver in filter(lambda p: ex.people_capacity[p] > 0,
range(ex.num_people)):
if assignments_dict[driver] == []:
assignment[driver] = []
total_cost += heuristics.pickup_cost(ex, driver,
assignments_dict[driver])[0]
assignment[driver] = heuristics.pickup_cost(
ex, driver, assignments_dict[driver])[1]
return (total_cost, assignment)
def agglomerative(ex):
"""
@param ex: Explorer
"""
row_names = []
col_names = []
people_with_cars = [i for i, x in enumerate(ex.people_capacity) if x > 0]
people_without_cars = [i for i, x in enumerate(ex.people_capacity) if x == 0]
# initiate matrix with non-car people on rows and car people on columns
first_i = -1
distance_array = []
assignments_dict = {}
# initialize dictionary of assignments_dict
# car_centroids = {}
for car_person in people_with_cars:
assignments_dict[car_person] = []
# car_centroids[car_person] = ex.locations[car_person]
# initialize matrix of distances between non-car people to car people
for i in people_with_cars:
person_dist = []
row_names.append(i)
if first_i == -1:
first_i = i
for j in people_without_cars:
if i == first_i:
col_names.append(j)
person_dist.append(ex.distances[i][j]["weight"])
distance_array.append(person_dist)
distance_matrix = np.matrix(distance_array)
if ex.verbose:
print "distance_matrix"
print distance_matrix
# while there is an unassigned non-car person, assign the person with minimum distance to a car with space. Then recenter the location of that car.
while len(col_names) > 0:
min_row = np.unravel_index(np.argmin(distance_matrix), np.shape(distance_matrix))[0]
min_col = np.unravel_index(np.argmin(distance_matrix), np.shape(distance_matrix))[1]
min_car = row_names[min_row]
assignments_dict[min_car].append(col_names[min_col])
distance_matrix = np.delete(distance_matrix, min_col, 1)
del col_names[min_col]
if len(assignments_dict[min_car]) >= (ex.people_capacity[min_car] - 1):
distance_matrix = np.delete(distance_matrix, min_row, 0)
del row_names[min_row]
# change the location of the car to be centroid of all people assigned to it and recompute distances for that row
else:
centroid_x, centroid_y = ex.locations[min_car]
for person in assignments_dict[min_car]:
centroid_x += ex.locations[person][0]
centroid_y += ex.locations[person][1]
centroid_x = centroid_x / (len(assignments_dict[min_car]) + 1)
centroid_y = centroid_y / (len(assignments_dict[min_car]) + 1)
new_row_dist = []
for person in col_names:
new_row_dist.append(
((ex.locations[person][0] - centroid_x) ** 2 + (ex.locations[person][1] - centroid_y) ** 2) ** 0.5
)
distance_matrix[min_row] = new_row_dist
total_cost = 0.0
assignment = [None] * ex.num_people
for driver in filter(lambda p: ex.people_capacity[p] > 0, range(ex.num_people)):
if assignments_dict[driver] == []:
assignment[driver] = []
total_cost += heuristics.pickup_cost(ex, driver, assignments_dict[driver])[0]
assignment[driver] = heuristics.pickup_cost(ex, driver, assignments_dict[driver])[1]
return (total_cost, assignment)
def kNearestDistance(ex):
"""
@param ex: Explorer
Algorithm based on projection distance
"""
### pick up passengers on the way ###
# create matrix (passenger x driver)
#projection_matrix = heuristics.get_passenger_driver_projection_matrix_ver2(ex, range(ex.num_people))
projection_matrix = heuristics.get_passenger_driver_distance_matrix(
ex, range(ex.num_people))
index_matrix = heuristics.get_passenger_driver_index_matrix(
ex, range(ex.num_people))
if ex.verbose:
print projection_matrix
print index_matrix
# for each global minimum projection, assign passenger to the car
# remove passenger row
# once the car is filled, remove car column
assignment = [None] * ex.num_people
while np.size(projection_matrix) > 0 and not np.isnan(
np.nanmin(projection_matrix)):
(num_unassign_passengers, num_avail_cars) = projection_matrix.shape
min_index = np.nanargmin(projection_matrix)
min_index_0 = min_index / num_avail_cars
min_index_1 = min_index % num_avail_cars
(min_passenger_idx,
min_car_idx) = index_matrix[min_index_0][min_index_1]
if ex.verbose:
print projection_matrix
print 'min: %s at (%s, %s)' % (str(np.nanmin(projection_matrix)),
str(min_index_0), str(min_index_1))
# assign min passenger to min car
list_min_car_passengers = assignment[min_car_idx]
if (list_min_car_passengers == None):
list_min_car_passengers = []
list_min_car_passengers.append(min_passenger_idx)
if len(list_min_car_passengers) <= ex.people_capacity[min_car_idx] - 1:
assignment[min_car_idx] = list_min_car_passengers
if ex.verbose:
print 'assign passenger %s to driver %s' % (
str(min_passenger_idx), str(min_car_idx))
# remove car column if car is full after taking this passenger
if (len(list_min_car_passengers) == ex.people_capacity[min_car_idx] -
1):
projection_matrix = np.delete(projection_matrix,
min_index_1,
axis=1)
index_matrix = np.delete(index_matrix, min_index_1, axis=1)
# remove passenger
projection_matrix = np.delete(projection_matrix, min_index_0, axis=0)
index_matrix = np.delete(index_matrix, min_index_0, axis=0)
### for remaining passengers, assign to closest driver ###
if np.size(projection_matrix) > 0:
# get a list of people indices
list_index_matrix = index_matrix.tolist()
list_index_cars = [tuple[1] for tuple in list_index_matrix[0]]
list_index_passengers = []
for list_index in list_index_matrix:
list_index_passengers.append(list_index[0][0])
list_people = list_index_cars + list_index_passengers
# get distance and index matrices
distance_matrix = heuristics.get_passenger_driver_distance_matrix(
ex, list_people)
index_matrix = heuristics.get_passenger_driver_index_matrix(
ex, list_people)
# for each global minimum distance, assign passenger to car
while np.size(distance_matrix) > 0 and not np.isnan(
np.nanmin(distance_matrix)):
(num_unassign_passengers, num_avail_cars) = distance_matrix.shape
min_index = np.nanargmin(distance_matrix)
min_index_0 = min_index / num_avail_cars
min_index_1 = min_index % num_avail_cars
(min_passenger_idx,
min_car_idx) = index_matrix[min_index_0][min_index_1]
# assign min passenger to min car
list_min_car_passengers = assignment[min_car_idx]
if (list_min_car_passengers == None):
list_min_car_passengers = []
list_min_car_passengers.append(min_passenger_idx)
if len(list_min_car_passengers
) <= ex.people_capacity[min_car_idx] - 1:
assignment[min_car_idx] = list_min_car_passengers
# remove car column if car is full after taking this passenger
if (len(list_min_car_passengers) == ex.people_capacity[min_car_idx]
- 1):
distance_matrix = np.delete(distance_matrix,
min_index_1,
axis=1)
index_matrix = np.delete(index_matrix, min_index_1, axis=1)
# remove passenger
distance_matrix = np.delete(distance_matrix, min_index_0, axis=0)
index_matrix = np.delete(index_matrix, min_index_0, axis=0)
if ex.verbose:
print 'assignment:'
print assignment
# calculate cost
total_cost = 0.0
for driver in filter(lambda p: ex.people_capacity[p] > 0,
range(ex.num_people)):
if assignment[driver] == None:
assignment[driver] = []
total_cost += heuristics.pickup_cost(ex, driver, assignment[driver])[0]
assignment[driver] = heuristics.pickup_cost(ex, driver,
assignment[driver])[1]
return (total_cost, assignment)
def kNearestDistance(ex):
"""
@param ex: Explorer
Algorithm based on projection distance
"""
### pick up passengers on the way ###
# create matrix (passenger x driver)
# projection_matrix = heuristics.get_passenger_driver_projection_matrix_ver2(ex, range(ex.num_people))
projection_matrix = heuristics.get_passenger_driver_distance_matrix(ex, range(ex.num_people))
index_matrix = heuristics.get_passenger_driver_index_matrix(ex, range(ex.num_people))
if ex.verbose:
print projection_matrix
print index_matrix
# for each global minimum projection, assign passenger to the car
# remove passenger row
# once the car is filled, remove car column
assignment = [None] * ex.num_people
while np.size(projection_matrix) > 0 and not np.isnan(np.nanmin(projection_matrix)):
(num_unassign_passengers, num_avail_cars) = projection_matrix.shape
min_index = np.nanargmin(projection_matrix)
min_index_0 = min_index / num_avail_cars
min_index_1 = min_index % num_avail_cars
(min_passenger_idx, min_car_idx) = index_matrix[min_index_0][min_index_1]
if ex.verbose:
print projection_matrix
print "min: %s at (%s, %s)" % (str(np.nanmin(projection_matrix)), str(min_index_0), str(min_index_1))
# assign min passenger to min car
list_min_car_passengers = assignment[min_car_idx]
if list_min_car_passengers == None:
list_min_car_passengers = []
list_min_car_passengers.append(min_passenger_idx)
if len(list_min_car_passengers) <= ex.people_capacity[min_car_idx] - 1:
assignment[min_car_idx] = list_min_car_passengers
if ex.verbose:
print "assign passenger %s to driver %s" % (str(min_passenger_idx), str(min_car_idx))
# remove car column if car is full after taking this passenger
if len(list_min_car_passengers) == ex.people_capacity[min_car_idx] - 1:
projection_matrix = np.delete(projection_matrix, min_index_1, axis=1)
index_matrix = np.delete(index_matrix, min_index_1, axis=1)
# remove passenger
projection_matrix = np.delete(projection_matrix, min_index_0, axis=0)
index_matrix = np.delete(index_matrix, min_index_0, axis=0)
### for remaining passengers, assign to closest driver ###
if np.size(projection_matrix) > 0:
# get a list of people indices
list_index_matrix = index_matrix.tolist()
list_index_cars = [tuple[1] for tuple in list_index_matrix[0]]
list_index_passengers = []
for list_index in list_index_matrix:
list_index_passengers.append(list_index[0][0])
list_people = list_index_cars + list_index_passengers
# get distance and index matrices
distance_matrix = heuristics.get_passenger_driver_distance_matrix(ex, list_people)
index_matrix = heuristics.get_passenger_driver_index_matrix(ex, list_people)
# for each global minimum distance, assign passenger to car
while np.size(distance_matrix) > 0 and not np.isnan(np.nanmin(distance_matrix)):
(num_unassign_passengers, num_avail_cars) = distance_matrix.shape
min_index = np.nanargmin(distance_matrix)
min_index_0 = min_index / num_avail_cars
min_index_1 = min_index % num_avail_cars
(min_passenger_idx, min_car_idx) = index_matrix[min_index_0][min_index_1]
# assign min passenger to min car
list_min_car_passengers = assignment[min_car_idx]
if list_min_car_passengers == None:
list_min_car_passengers = []
list_min_car_passengers.append(min_passenger_idx)
if len(list_min_car_passengers) <= ex.people_capacity[min_car_idx] - 1:
assignment[min_car_idx] = list_min_car_passengers
# remove car column if car is full after taking this passenger
if len(list_min_car_passengers) == ex.people_capacity[min_car_idx] - 1:
distance_matrix = np.delete(distance_matrix, min_index_1, axis=1)
index_matrix = np.delete(index_matrix, min_index_1, axis=1)
# remove passenger
distance_matrix = np.delete(distance_matrix, min_index_0, axis=0)
index_matrix = np.delete(index_matrix, min_index_0, axis=0)
if ex.verbose:
print "assignment:"
print assignment
# calculate cost
total_cost = 0.0
for driver in filter(lambda p: ex.people_capacity[p] > 0, range(ex.num_people)):
if assignment[driver] == None:
assignment[driver] = []
total_cost += heuristics.pickup_cost(ex, driver, assignment[driver])[0]
assignment[driver] = heuristics.pickup_cost(ex, driver, assignment[driver])[1]
return (total_cost, assignment)