-
Notifications
You must be signed in to change notification settings - Fork 16
/
CVRPTWdata_generate.py
169 lines (152 loc) · 7.82 KB
/
CVRPTWdata_generate.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
import os
from subprocess import check_call
from multiprocessing import Pool
import tqdm
import numpy as np
import pickle
import torch
from torch.autograd import Variable
from tqdm import trange
import argparse
import time
import tempfile
def write_instance(instance, instance_name, instance_filename):
with open(instance_filename, "w") as f:
x = instance[0]
demand = instance[1]
capacity = instance[2]
a = instance[3]
b = instance[4]
service_time = instance[5]
n_nodes = x.shape[0]
f.write("NAME : " + instance_name + "\n")
f.write("COMMENT : blank\n")
f.write("TYPE : CVRPTW\n")
f.write("VEHICLES : 20\n")
f.write("CAPACITY : " + str(capacity) + "\n")
f.write("SERVICE_TIME : " + str(service_time * 1000000) + "\n" )
f.write("DIMENSION : " + str(n_nodes) + "\nEDGE_WEIGHT_TYPE : EUC_2D\n")
f.write("NODE_COORD_SECTION\n")
for l in range(n_nodes):
f.write(" "+str(l+1)+" "+str(x[l][0]*1000000)[:15]+" "+str(x[l][1]*1000000)[:15]+"\n")
f.write("DEMAND_SECTION\n")
f.write("1 0\n")
for l in range(n_nodes - 1):
f.write(str(l + 2) + " " + str(int(demand[l]))+"\n")
f.write("TIME_WINDOW_SECTION\n")
f.write("1 0 10000000\n")
for l in range(n_nodes - 1):
f.write(str(l + 2) + " " + str(int(a[l] * 1000000)) + " " + str(int(b[l] * 1000000)) + "\n")
f.write("DEPOT_SECTION\n 1\n -1\n")
f.write("EOF\n")
def write_para(dataset_name, instance_name, instance_filename, method, para_filename, max_trials=1000, seed=1234):
with open(para_filename, "w") as f:
f.write("PROBLEM_FILE = " + instance_filename + "\n")
f.write("MAX_TRIALS = " + str(max_trials) + "\n")
f.write("SPECIAL\nRUNS = 1\n")
f.write("SEED = " + str(seed) + "\n")
def method_wrapper(args):
if args[0] == "LKH":
return solve_LKH(*args[1:])
elif args[0] == "FeatGen":
return generate_feat(*args[1:])
def solve_LKH(dataset_name, instance, instance_name, rerun=False, max_trials=1000):
para_filename = "tmp/" + dataset_name + "/LKH_para/" + instance_name + ".para"
log_filename = "tmp/" + dataset_name + "/LKH_log/" + instance_name + ".log"
instance_filename = "tmp/" + dataset_name + "/cvrptw/" + instance_name + ".cvrptw"
if rerun or not os.path.isfile(log_filename):
write_instance(instance, instance_name, instance_filename)
write_para(dataset_name, instance_name, instance_filename, "LKH", para_filename, max_trials=max_trials)
with open(log_filename, "w") as f:
check_call(["./LKH", para_filename], stdout=f)
return read_results(log_filename, max_trials)
def read_results(log_filename, max_trials):
with open(log_filename, "r") as f:
line = f.readlines()[-1]
result = [int(_line) - 1 for _line in line.split(" ")[:-2]]
return result
def generate_dataset(n_samples, n_nodes, save_dir):
capacity = 1000
service_time = 0.1
loc = np.random.uniform(size=(n_samples, n_nodes + 1, 2))
demand = np.random.normal(15, 10, (n_samples, n_nodes)).astype("int")
demand = np.maximum(np.minimum(np.ceil(np.abs(demand)), 42), 1)
dist = np.sqrt(((loc[:, 0:1] - loc[:, 1:]) ** 2).sum(-1)) * 100
a_sample = np.floor(dist) + 1
b_sample = 1000 - a_sample - 10
a = np.random.uniform(size=(n_samples, n_nodes))
a = (a * (b_sample - a_sample) + a_sample).astype("int")
eps = np.maximum(np.abs(np.random.normal(0, 1, (n_samples, n_nodes))), 0.01)
b = np.minimum(np.ceil(a + 300 * eps), b_sample)
a = a / 100
b = b / 100
dataset = {"loc":loc,
"demand":demand,
"start":a,
"end":b,
"capacity":capacity,
"service_time":service_time}
if save_dir == "CVRPTW_test":
with open(save_dir + "/cvrptw_" + str(n_nodes) + ".pkl", "wb") as f:
pickle.dump(dataset, f)
return
data = [[dataset["loc"][i], dataset["demand"][i], dataset["capacity"], dataset["start"][i], dataset["end"][i], dataset["service_time"]] for i in range(n_samples)]
n_neighbours = 20
os.makedirs("tmp/" + str(n_nodes) + "/cvrptw", exist_ok=True)
os.makedirs("tmp/" + str(n_nodes) + "/LKH_para", exist_ok=True)
os.makedirs("tmp/" + str(n_nodes) + "/LKH_log", exist_ok=True)
demand = np.concatenate([np.zeros((n_samples, 1)), dataset['demand'] / 50], -1)
start = np.concatenate([np.zeros((n_samples, 1)), dataset['start'] / 10], -1)
end = np.concatenate([np.ones((n_samples, 1)), dataset['end'] / 10], -1)
capacity = np.concatenate([np.ones((n_samples, 1)), np.zeros((n_samples, n_nodes))], -1)
x = dataset['loc']
node_feat = np.concatenate([x,
demand.reshape(n_samples, n_nodes + 1, 1),
start.reshape(n_samples, n_nodes + 1, 1),
end.reshape(n_samples, n_nodes + 1, 1),
capacity.reshape(n_samples, n_nodes + 1, 1)], -1)
dist = x.reshape(n_samples, n_nodes + 1, 1, 2) - x.reshape(n_samples, 1, n_nodes + 1, 2)
dist = np.sqrt((dist ** 2).sum(-1)) # 10000 x 100 x 100
edge_index = np.argsort(dist, -1)[:, :, 1:1 + n_neighbours]
edge_feat = dist[np.arange(n_samples).reshape(-1, 1, 1), np.arange(n_nodes + 1).reshape(1, -1, 1), edge_index]
inverse_edge_index = -np.ones(shape=[n_samples, n_nodes + 1, n_nodes + 1], dtype="int")
inverse_edge_index[np.arange(n_samples).reshape(-1, 1, 1), edge_index, np.arange(n_nodes + 1).reshape(1, -1, 1)] = np.arange(n_neighbours).reshape(1, 1, -1) + np.arange(n_nodes + 1).reshape(1, -1, 1) * n_neighbours
inverse_edge_index = inverse_edge_index[np.arange(n_samples).reshape(-1, 1, 1), np.arange(n_nodes + 1).reshape(1, -1, 1), edge_index]
with Pool(os.cpu_count()) as pool:
result = list(tqdm.tqdm(pool.imap(method_wrapper, [("LKH", str(n_nodes), data[i], str(i), True, 10000) for i in range(len(data))]), total=len(data)))
result = np.array(result) # n_samples x n_nodes
result[result > n_nodes] = 0
label1 = np.zeros(shape=[n_samples, n_nodes + 1, n_nodes + 1], dtype="bool")
label2 = np.zeros(shape=[n_samples, n_nodes + 1, n_nodes + 1], dtype="bool")
label1[np.arange(n_samples).reshape(-1, 1), result, np.roll(result, 1, -1)] = True
label2[np.arange(n_samples).reshape(-1, 1), np.roll(result, 1, -1), result] = True
label1 = label1[np.arange(n_samples).reshape(-1, 1, 1), np.arange(n_nodes + 1).reshape(1, -1, 1), edge_index]
label2 = label2[np.arange(n_samples).reshape(-1, 1, 1), np.arange(n_nodes + 1).reshape(1, -1, 1), edge_index]
feat = {"node_feat":node_feat,
"edge_feat":edge_feat,
"edge_index":edge_index,
"inverse_edge_index":inverse_edge_index,
"label1":label1,
"label2":label2}
with open(save_dir + "/" + str(n_nodes) + ".pkl", "wb") as f:
pickle.dump(feat, f)
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='')
parser.add_argument("-train", action='store_true', help="Generate training and validation datasets")
parser.add_argument("-test", action='store_true', help="Generate test datasets")
args = parser.parse_args()
if args.train:
os.makedirs("CVRPTW_train", exist_ok=True)
os.makedirs("CVRPTW_val", exist_ok=True)
for n_nodes in range(41, 201):
n_samples = 2 * 120000 // n_nodes
generate_dataset(n_samples, n_nodes, "CVRPTW_train")
for n_nodes in [40, 80, 200]:
n_samples = 1000
generate_dataset(n_samples, n_nodes, "CVRPTW_val")
if args.test:
os.makedirs("CVRPTW_test", exist_ok=True)
n_samples = 1000
for n_nodes in [40, 200, 300]:
np.random.seed(1234)
generate_dataset(n_samples, n_nodes, "CVRPTW_test")