def test_calculate_cost_calculator_with_one_cars():
solution = Solution(vehicle_routes=np.array([[2, 1, 0], [0, 0, 0]]))
node_distances = NodeDistances(
distances=np.array([[0, 1, 2], [3, 0, 5], [6, 7, 0]]))
calculator = SolutionRestrictionsCalculator(node_distances=node_distances,
vehicle_allowed_distances=None)
cost = calculator.calculate_cost(solution=solution)
assert 12 == cost, "All only one route should sum 12"
class SolutionInitializer(object):
"""
Initializes a solution.
"""
def __init__(self, node_distances: NodeDistances,
vehicle_allowed_distances: VehicleAllowedDistances):
self.node_distances = node_distances
self.vehicle_allowed_distances = vehicle_allowed_distances
self.solution_restrictions_calculator=SolutionRestrictionsCalculator(node_distances=node_distances, \
vehicle_allowed_distances=vehicle_allowed_distances)
def init_randomly(self) -> Solution:
"""
For each customer we randomly select a vehicle. If the fleet contains vehicles which cannot performe a certain
trip, it may lead to unfeasible solutions.
We reject invalid solutions, so we will try till a valid ones is obtained randomly.
"""
return self.__init_solution(
initializing_function=self.__assign_nodes_randomly)
def init_one_node_to_one_vehicle(self) -> Solution:
"""
We assign randomly one node to one vehicle.
This should lead to a valid solution. However we will check it and look for another solution in case we found
a no-valid one.
"""
return self.__init_solution(
initializing_function=self.__assign_one_node_to_one_vehicle)
def init_vehicles_with_group_nodes_sequentially(self) -> Solution:
"""
We assign nodes sequeantially to vehicles till they reach their restriction.
This should lead to a valid solution. However we will check it and look for another solution in case we found
a no-valid one.
"""
return self.__init_solution(
initializing_function=self.
__assign_group_nodes_sequentially_to_vehicles)
def __init_solution(self, initializing_function) -> Solution:
"""
Initializing the solution, using the function passed as parameter to create the vehicle routes.
After 100 attempts we consider an error must be happening and a exception is raised.
"""
attempt = 0
solution_is_valid = False
while solution_is_valid == False:
if attempt >= 100:
raise Exception("Initial VALID solution not found after " +
str(attempt) + " attempts.")
attempt = attempt + 1
vehicle_routes = initializing_function()
solution = Solution(vehicle_routes=vehicle_routes)
solution_is_valid = self.solution_restrictions_calculator.check_if_solution_is_valid(
solution=solution)
solution.is_valid = True
solution.cost = self.solution_restrictions_calculator.calculate_cost(
solution=solution)
return solution
def __assign_nodes_randomly(self) -> np.ndarray:
"""
Assign nodes randomly
"""
num_vehicles = self.vehicle_allowed_distances.get_number_of_vehicles()
num_nodes = self.node_distances.get_number_of_nodes()
vehicle_routes = np.zeros([num_vehicles, num_nodes], dtype=int)
for node in range(1, num_nodes):
vehicle = random.randint(0, num_vehicles - 1)
route = vehicle_routes[vehicle]
available_indexes = np.where(route == 0)[0]
index = available_indexes[0]
vehicle_routes[vehicle][index] = node
return vehicle_routes
def __assign_group_nodes_sequentially_to_vehicles(self) -> np.ndarray:
"""
We assign nodes sequeantially to vehicles till they reach their restriction.
This will lead to a valid solution.
"""
allowed_distances = self.vehicle_allowed_distances.distances
num_nodes = self.node_distances.get_number_of_nodes()
vehicle_routes = np.zeros([len(allowed_distances), num_nodes],
dtype=int)
vehicle = 0
index_node = 0
node_list = list(range(1, num_nodes))
random.shuffle(node_list)
for node in node_list: #we suffle from 1 in order to ignore depot (=0)
#Asign and check if valid
vehicle_routes[vehicle][index_node] = node
remaining_distance=self.solution_restrictions_calculator.get_remaining_distance(route=vehicle_routes[vehicle], \
node_distances=self.node_distances, allowed_distance=allowed_distances[vehicle])
#If no valid, undo the assignmment and move to next vehicle.
if remaining_distance < 0:
vehicle_routes[vehicle][index_node] = 0
vehicle = vehicle + 1
index_node = 0
vehicle_routes[vehicle][index_node] = node
#Move to next index.
index_node = index_node + 1
return vehicle_routes
def __assign_one_node_to_one_vehicle(self) -> np.ndarray:
"""
As Clark and Wright, we assign one node to one vehicle.
"""
num_vehicles = self.vehicle_allowed_distances.get_number_of_vehicles()
num_nodes = self.node_distances.get_number_of_nodes()
vehicle_routes = np.zeros([num_vehicles, num_nodes], dtype=int)
suffled_customers = np.array(range(1, num_nodes), dtype=int)
random.shuffle(suffled_customers)
for index in range(0, num_vehicles):
vehicle_routes[index][0] = suffled_customers[index]
return vehicle_routes