-
Notifications
You must be signed in to change notification settings - Fork 9
/
aco_tsp.py
203 lines (180 loc) · 9.11 KB
/
aco_tsp.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
import math
import random
import time as tm
from tqdm import tqdm
from matplotlib import pyplot as plt
class SolveTSPUsingACO:
class Edge:
def __init__(self, a, b, weight, initial_pheromone):
self.a = a
self.b = b
if weight == 0:
weight = 1e-10
self.weight = weight
self.pheromone = initial_pheromone
class Ant:
def __init__(self, alpha, beta, num_nodes, edges):
self.alpha = alpha
self.beta = beta
self.num_nodes = num_nodes
self.edges = edges
self.tour = None
self.distance = 0.0
def _select_node(self):
roulette_wheel = 0.0
unvisited_nodes = [node for node in range(self.num_nodes) if node not in self.tour]
heuristic_total = 0.0
for unvisited_node in unvisited_nodes:
heuristic_total += self.edges[self.tour[-1]][unvisited_node].weight
for unvisited_node in unvisited_nodes:
roulette_wheel += pow(self.edges[self.tour[-1]][unvisited_node].pheromone, self.alpha) * \
pow((heuristic_total / self.edges[self.tour[-1]][unvisited_node].weight), self.beta)
random_value = random.uniform(0.0, roulette_wheel)
wheel_position = 0.0
for unvisited_node in unvisited_nodes:
wheel_position += pow(self.edges[self.tour[-1]][unvisited_node].pheromone, self.alpha) * \
pow((heuristic_total / self.edges[self.tour[-1]][unvisited_node].weight), self.beta)
if wheel_position >= random_value:
return unvisited_node
def find_tour(self):
self.tour = [random.randint(0, self.num_nodes - 1)]
while len(self.tour) < self.num_nodes:
self.tour.append(self._select_node())
return self.tour
def get_distance(self):
self.distance = 0.0
for i in range(self.num_nodes):
self.distance += self.edges[self.tour[i]][self.tour[(i + 1) % self.num_nodes]].weight
return self.distance
def __init__(self, mode='ACS', colony_size=10, elitist_weight=1.0, min_scaling_factor=0.001, alpha=1.0, beta=3.0,
rho=0.1, pheromone_deposit_weight=1.0, initial_pheromone=1.0, steps=100, nodes=None, labels=None):
self.mode = mode
self.colony_size = colony_size
self.elitist_weight = elitist_weight
self.min_scaling_factor = min_scaling_factor
self.rho = rho
self.pheromone_deposit_weight = pheromone_deposit_weight
self.steps = steps
self.num_nodes = len(nodes)
self.nodes = nodes
if labels is not None:
self.labels = labels
else:
self.labels = range(1, self.num_nodes + 1)
self.edges = [[None] * self.num_nodes for _ in range(self.num_nodes)]
for i in range(self.num_nodes):
for j in range(i + 1, self.num_nodes):
self.edges[i][j] = self.edges[j][i] = self.Edge(i, j, math.sqrt(
pow(self.nodes[i][0] - self.nodes[j][0], 2.0) + pow(self.nodes[i][1] - self.nodes[j][1], 2.0)),
initial_pheromone)
self.ants = [self.Ant(alpha, beta, self.num_nodes, self.edges) for _ in range(self.colony_size)]
self.global_best_tour = None
self.global_best_distance = float("inf")
def _add_pheromone(self, tour, distance, weight=1.0):
pheromone_to_add = self.pheromone_deposit_weight / distance
for i in range(self.num_nodes):
self.edges[tour[i]][tour[(i + 1) % self.num_nodes]].pheromone += weight * pheromone_to_add
def _acs(self):
for step in range(self.steps):
for ant in self.ants:
# Amout of pheromone increment is constant.
self._add_pheromone(ant.find_tour(), ant.get_distance())
if ant.distance < self.global_best_distance:
self.global_best_tour = ant.tour
self.global_best_distance = ant.distance
for i in range(self.num_nodes):
for j in range(i + 1, self.num_nodes):
self.edges[i][j].pheromone *= (1.0 - self.rho)
def _elitist(self):
for step in range(self.steps):
for ant in self.ants:
self._add_pheromone(ant.find_tour(), ant.get_distance())
if ant.distance < self.global_best_distance:
self.global_best_tour = ant.tour
self.global_best_distance = ant.distance
# Amout of pheromone depends upon the quality of the solution.
self._add_pheromone(self.global_best_tour, self.global_best_distance, weight=self.elitist_weight)
for i in range(self.num_nodes):
for j in range(i + 1, self.num_nodes):
self.edges[i][j].pheromone *= (1.0 - self.rho)
def _max_min(self):
for step in range(self.steps):
iteration_best_tour = None
iteration_best_distance = float("inf")
for ant in self.ants:
ant.find_tour()
if ant.get_distance() < iteration_best_distance:
iteration_best_tour = ant.tour
iteration_best_distance = ant.distance
if float(step + 1) / float(self.steps) <= 0.75:
# Initially higher weight gradually reduced will 75% completion
self._add_pheromone(iteration_best_tour, iteration_best_distance)
max_pheromone = self.pheromone_deposit_weight / iteration_best_distance
else:
# Last 25% weights based on quality of route after comparing with the global best tour.
if iteration_best_distance < self.global_best_distance:
self.global_best_tour = iteration_best_tour
self.global_best_distance = iteration_best_distance
self._add_pheromone(self.global_best_tour, self.global_best_distance)
max_pheromone = self.pheromone_deposit_weight / self.global_best_distance
min_pheromone = max_pheromone * self.min_scaling_factor
# Setting edge weights to min and max phermones
for i in range(self.num_nodes):
for j in range(i + 1, self.num_nodes):
self.edges[i][j].pheromone *= (1.0 - self.rho)
if self.edges[i][j].pheromone > max_pheromone:
self.edges[i][j].pheromone = max_pheromone
elif self.edges[i][j].pheromone < min_pheromone:
self.edges[i][j].pheromone = min_pheromone
def run(self):
start = tm.time()
# print('Started : {0}'.format(self.mode))
if self.mode == 'ACS':
self._acs()
elif self.mode == 'Elitist':
self._elitist()
else:
self._max_min()
# print('Ended : {0}'.format(self.mode))
# print('Sequence : <- {0} ->'.format(' - '.join(str(self.labels[i]) for i in self.global_best_tour)))
# print('Total distance travelled to complete the tour : {0}\n'.format(round(self.global_best_distance, 2)))
runtime = tm.time() - start
return runtime, self.global_best_distance
def plot(self, line_width=1, point_radius=math.sqrt(2.0), annotation_size=8, dpi=120, save=True, name=None):
x = [self.nodes[i][0] for i in self.global_best_tour]
x.append(x[0])
y = [self.nodes[i][1] for i in self.global_best_tour]
y.append(y[0])
plt.plot(x, y, linewidth=line_width)
plt.scatter(x, y, s=math.pi * (point_radius ** 2.0))
plt.title(self.mode)
for i in self.global_best_tour:
plt.annotate(self.labels[i], self.nodes[i], size=annotation_size)
if save:
if name is None:
name = '{0}_tour.png'.format(self.mode)
plt.savefig(name, dpi=dpi)
plt.show()
plt.gcf().clear()
if __name__ == '__main__':
_colony_size = 5
_steps = 50
f = open('./out.csv', 'w')
f.write(','.join(['Iteration', 'ACS_time', 'ACS_dist', 'Elitist_time', \
'Elitist_dist', 'MaxMin_time', 'MaxMin_dist']))
f.write('\n')
# Analyse and compare the various modes
for i in range(10):
print('Iter: ', i+1)
for j in tqdm(range(20)):
_nodes = [(random.uniform(-400, 400), random.uniform(-400, 400)) for _ in range(0, 10*(i+1))]
acs = SolveTSPUsingACO(mode='ACS', colony_size=_colony_size, steps=_steps, nodes=_nodes)
elitist = SolveTSPUsingACO(mode='Elitist', colony_size=_colony_size, steps=_steps, nodes=_nodes)
max_min = SolveTSPUsingACO(mode='MaxMin', colony_size=_colony_size, steps=_steps, nodes=_nodes)
time, dist = acs.run()
f.write(str(i+1) + "," + str(time) + "," + str(dist) + ",")
time, dist = elitist.run()
f.write(str(time) + "," + str(dist) + ",")
time, dist = max_min.run()
f.write(str(time) + "," + str(dist) + "\n")
f.close()