用Python定义旅行推销员的线性规划模型
利用Python和用Python定义旅行推销员的线性规划模型,python,algorithm,python-2.7,linear-programming,pulp,Python,Algorithm,Python 2.7,Linear Programming,Pulp,利用Python和pill库,我们如何创建线性规划模型来解决旅行商问题(TSP) 在维基百科中,目标函数和约束条件是 问题:这是我的部分尝试,我被卡住了 我没有在代码中包含最终约束,因为我不知道如何定义它。我相信u变量的约束是为了防止解决方案中出现子循环 此外,对当前模型的求解给出了决策变量,如x0_0和x1_1等于1.0,这肯定是错误的。。。我不明白为什么会这样即使我有 if i == j: upperBound = 0 Python代码 import pulp
pill
库,我们如何创建线性规划模型来解决旅行商问题(TSP)
在维基百科中,目标函数和约束条件是
问题:这是我的部分尝试,我被卡住了
x0_0
和x1_1
等于1.0
,这肯定是错误的。。。我不明白为什么会这样即使我有
if i == j:
upperBound = 0
import pulp
def get_dist(tsp):
with open(tsp, 'rb') as tspfile:
r = csv.reader(tspfile, delimiter='\t')
d = [row for row in r]
d = d[1:] # skip header row
locs = set([r[0] for r in d]) | set([r[1] for r in d])
loc_map = {l:i for i, l in enumerate(locs)}
idx_map = {i:l for i, l in enumerate(locs)}
dist = [(loc_map[r[0]], loc_map[r[1]], r[2]) for r in d]
return dist, idx_map
def dist_from_coords(dist, n):
points = []
for i in range(n):
points.append([0] * n)
for i, j, v in dist:
points[i][j] = points[j][i] = float(v)
return points
def find_tour():
tsp_file = '/Users/test/' + 'my-waypoints-dist-dur.tsv'
coords, idx_map = get_dist(tsp_file)
n = len(idx_map)
dist = dist_from_coords(coords, n)
# Define the problem
m = pulp.LpProblem('TSP', pulp.LpMinimize)
# Create variables
# x[i,j] is 1 if edge i->j is on the optimal tour, and 0 otherwise
# Also forbid loops
x = {}
for i in range(n):
for j in range(n):
lowerBound = 0
upperBound = 1
# Forbid loops
if i == j:
upperBound = 0
# print i,i
x[i,j] = pulp.LpVariable('x' + str(i) + '_' + str(j), lowerBound, upperBound, pulp.LpBinary)
# x[j,i] = x[i,j]
# Define the objective function to minimize
m += pulp.lpSum([dist[i][j] * x[i,j] for i in range(n) for j in range(n)])
# Add degree-2 constraint
for i in range(n):
m += pulp.lpSum([x[i,j] for j in range(n)]) == 2
# Solve and display results
status = m.solve()
print pulp.LpStatus[status]
for i in range(n):
for j in range(n):
if pulp.value(x[i,j]) >0:
print str(i) + '_' + str(j) + ': ' + str( pulp.value(x[i,j]) )
find_tour()
import csv
import pulp
def get_dist(tsp):
with open(tsp, 'rb') as tspfile:
r = csv.reader(tspfile, delimiter='\t')
d = [row for row in r]
d = d[1:] # skip header row
locs = set([r[0] for r in d]) | set([r[1] for r in d])
loc_map = {l:i for i, l in enumerate(locs)}
idx_map = {i:l for i, l in enumerate(locs)}
dist = [(loc_map[r[0]], loc_map[r[1]], r[2]) for r in d]
return dist, idx_map
def dist_from_coords(dist, n):
points = []
for i in range(n):
points.append([0] * n)
for i, j, v in dist:
points[i][j] = points[j][i] = float(v)
return points
def find_tour():
tsp_file = '/Users/test/' + 'my-waypoints-dist-dur.tsv'
coords, idx_map = get_dist(tsp_file)
n = len(idx_map)
dist = dist_from_coords(coords, n)
# Define the problem
m = pulp.LpProblem('TSP', pulp.LpMinimize)
# Create variables
# x[i,j] is 1 if edge i->j is on the optimal tour, and 0 otherwise
# Also forbid loops
x = {}
for i in range(n+1):
for j in range(n+1):
lowerBound = 0
upperBound = 1
# Forbid loops
if i == j:
upperBound = 0
# print i,i
# Create the decision variable and First constraint
x[i,j] = pulp.LpVariable('x' + str(i) + '_' + str(j), lowerBound, upperBound, pulp.LpBinary)
# x[j,i] = x[i,j]
# Define the objective function to minimize
m += pulp.lpSum([dist[i][j] * x[i,j] for i in range(1,n+1) for j in range(1,n+1)])
# Add degree-2 constraint (3rd and 4th)
for i in range(1,n+1):
m += pulp.lpSum([x[i,j] for j in range(1,n+1)]) == 2
# Add the last (5th) constraint (prevents subtours)
u = []
for i in range(1, n+1):
u.append(pulp.LpVariable('u_' + str(i), cat='Integer'))
for i in range(1, n-1):
for j in range(i+1, n+1):
m += pulp.lpSum([ u[i] - u[j] + n*x[i,j]]) <= n-1
# status = m.solve()
# print pulp.LpStatus[status]
# for i in range(n):
# for j in range(n):
# if pulp.value(x[i,j]) >0:
# print str(i) + '_' + str(j) + ': ' + str( pulp.value(x[i,j]) )
find_tour()
tsv期间我的航路点距离
数据文件
结果
0_0: 1.0
0_5: 1.0
1_1: 1.0
1_15: 1.0
2_2: 1.0
2_39: 1.0
3_3: 1.0
3_26: 1.0
4_4: 1.0
4_42: 1.0
5_5: 1.0
5_33: 1.0
6_6: 1.0
6_31: 1.0
7_7: 1.0
7_38: 1.0
8_8: 1.0
8_24: 1.0
9_9: 1.0
9_26: 1.0
10_4: 1.0
10_10: 1.0
11_11: 1.0
11_12: 1.0
12_11: 1.0
12_12: 1.0
13_13: 1.0
13_17: 1.0
14_14: 1.0
14_18: 1.0
15_1: 1.0
15_15: 1.0
16_3: 1.0
16_16: 1.0
17_13: 1.0
17_17: 1.0
18_14: 1.0
18_18: 1.0
19_19: 1.0
19_20: 1.0
20_4: 1.0
20_20: 1.0
21_21: 1.0
21_25: 1.0
22_22: 1.0
22_27: 1.0
23_21: 1.0
23_23: 1.0
24_8: 1.0
24_24: 1.0
25_21: 1.0
25_25: 1.0
26_26: 1.0
26_43: 1.0
27_27: 1.0
27_38: 1.0
28_28: 1.0
28_47: 1.0
29_29: 1.0
29_31: 1.0
30_30: 1.0
30_34: 1.0
31_29: 1.0
31_31: 1.0
32_25: 1.0
32_32: 1.0
33_28: 1.0
33_33: 1.0
34_30: 1.0
34_34: 1.0
35_35: 1.0
35_42: 1.0
36_36: 1.0
36_47: 1.0
37_36: 1.0
37_37: 1.0
38_27: 1.0
38_38: 1.0
39_39: 1.0
39_44: 1.0
40_40: 1.0
40_43: 1.0
41_41: 1.0
41_45: 1.0
42_4: 1.0
42_42: 1.0
43_26: 1.0
43_43: 1.0
44_39: 1.0
44_44: 1.0
45_15: 1.0
45_45: 1.0
46_40: 1.0
46_46: 1.0
47_28: 1.0
47_47: 1.0
...
更新代码
import pulp
def get_dist(tsp):
with open(tsp, 'rb') as tspfile:
r = csv.reader(tspfile, delimiter='\t')
d = [row for row in r]
d = d[1:] # skip header row
locs = set([r[0] for r in d]) | set([r[1] for r in d])
loc_map = {l:i for i, l in enumerate(locs)}
idx_map = {i:l for i, l in enumerate(locs)}
dist = [(loc_map[r[0]], loc_map[r[1]], r[2]) for r in d]
return dist, idx_map
def dist_from_coords(dist, n):
points = []
for i in range(n):
points.append([0] * n)
for i, j, v in dist:
points[i][j] = points[j][i] = float(v)
return points
def find_tour():
tsp_file = '/Users/test/' + 'my-waypoints-dist-dur.tsv'
coords, idx_map = get_dist(tsp_file)
n = len(idx_map)
dist = dist_from_coords(coords, n)
# Define the problem
m = pulp.LpProblem('TSP', pulp.LpMinimize)
# Create variables
# x[i,j] is 1 if edge i->j is on the optimal tour, and 0 otherwise
# Also forbid loops
x = {}
for i in range(n):
for j in range(n):
lowerBound = 0
upperBound = 1
# Forbid loops
if i == j:
upperBound = 0
# print i,i
x[i,j] = pulp.LpVariable('x' + str(i) + '_' + str(j), lowerBound, upperBound, pulp.LpBinary)
# x[j,i] = x[i,j]
# Define the objective function to minimize
m += pulp.lpSum([dist[i][j] * x[i,j] for i in range(n) for j in range(n)])
# Add degree-2 constraint
for i in range(n):
m += pulp.lpSum([x[i,j] for j in range(n)]) == 2
# Solve and display results
status = m.solve()
print pulp.LpStatus[status]
for i in range(n):
for j in range(n):
if pulp.value(x[i,j]) >0:
print str(i) + '_' + str(j) + ': ' + str( pulp.value(x[i,j]) )
find_tour()
import csv
import pulp
def get_dist(tsp):
with open(tsp, 'rb') as tspfile:
r = csv.reader(tspfile, delimiter='\t')
d = [row for row in r]
d = d[1:] # skip header row
locs = set([r[0] for r in d]) | set([r[1] for r in d])
loc_map = {l:i for i, l in enumerate(locs)}
idx_map = {i:l for i, l in enumerate(locs)}
dist = [(loc_map[r[0]], loc_map[r[1]], r[2]) for r in d]
return dist, idx_map
def dist_from_coords(dist, n):
points = []
for i in range(n):
points.append([0] * n)
for i, j, v in dist:
points[i][j] = points[j][i] = float(v)
return points
def find_tour():
tsp_file = '/Users/test/' + 'my-waypoints-dist-dur.tsv'
coords, idx_map = get_dist(tsp_file)
n = len(idx_map)
dist = dist_from_coords(coords, n)
# Define the problem
m = pulp.LpProblem('TSP', pulp.LpMinimize)
# Create variables
# x[i,j] is 1 if edge i->j is on the optimal tour, and 0 otherwise
# Also forbid loops
x = {}
for i in range(n+1):
for j in range(n+1):
lowerBound = 0
upperBound = 1
# Forbid loops
if i == j:
upperBound = 0
# print i,i
# Create the decision variable and First constraint
x[i,j] = pulp.LpVariable('x' + str(i) + '_' + str(j), lowerBound, upperBound, pulp.LpBinary)
# x[j,i] = x[i,j]
# Define the objective function to minimize
m += pulp.lpSum([dist[i][j] * x[i,j] for i in range(1,n+1) for j in range(1,n+1)])
# Add degree-2 constraint (3rd and 4th)
for i in range(1,n+1):
m += pulp.lpSum([x[i,j] for j in range(1,n+1)]) == 2
# Add the last (5th) constraint (prevents subtours)
u = []
for i in range(1, n+1):
u.append(pulp.LpVariable('u_' + str(i), cat='Integer'))
for i in range(1, n-1):
for j in range(i+1, n+1):
m += pulp.lpSum([ u[i] - u[j] + n*x[i,j]]) <= n-1
# status = m.solve()
# print pulp.LpStatus[status]
# for i in range(n):
# for j in range(n):
# if pulp.value(x[i,j]) >0:
# print str(i) + '_' + str(j) + ': ' + str( pulp.value(x[i,j]) )
find_tour()
导入csv
进口纸浆
def get_dist(tsp):
打开(tsp,'rb')作为tsp文件:
r=csv.reader(tspfile,分隔符='\t')
d=[r中的行对行]
d=d[1:]#跳过标题行
locs=集合([r[0]表示d中的r])|集合([r[1]表示d中的r])
loc_map={l:i代表i,l在枚举(locs)}
idx_-map={i:l代表i,l在枚举(locs)}
对于d中的r,dist=[(loc_map[r[0]],loc_map[r[1]],r[2])]
返回距离,idx\U映射
def dist_from_coords(北区):
点数=[]
对于范围(n)中的i:
点。追加([0]*n)
对于距离中的i、j、v:
点[i][j]=点[j][i]=浮动(v)
返回点
def find_tour():
tsp_file='/Users/test/'+'my waypoints dist dur.tsv'
coords,idx\u map=get\u dist(tsp\u文件)
n=len(idx_映射)
距离=距离坐标的距离(坐标,n)
#定义问题
m=纸浆.lp问题('TSP',纸浆.lp最小化)
#创建变量
#如果边i->j处于最佳行程,则x[i,j]为1,否则为0
#也禁止循环
x={}
对于范围(n+1)内的i:
对于范围(n+1)内的j:
lowerBound=0
上限=1
#禁止循环
如果i==j:
上限=0
#打印i,i
#创建决策变量和第一个约束
x[i,j]=palp.LpVariable('x'+str(i)+'.'+str(j),下界,上界,palp.LpBinary)
#x[j,i]=x[i,j]
#定义最小化的目标函数
m+=纸浆lpSum([dist[i][j]*x[i,j]表示范围(1,n+1)内的i,表示范围(1,n+1)内的j)])
#添加阶数为2的约束(第3和第4)
对于范围(1,n+1)内的i:
m+=纸浆.lpSum([x[i,j]表示范围(1,n+1)内的j)]==2
#添加最后一个(第5个)约束(防止子任务)
u=[]
对于范围(1,n+1)内的i:
u、 追加(pulp.LpVariable('u_'+str(i),cat='Integer'))
对于范围(1,n-1)内的i:
对于范围(i+1,n+1)内的j:
m+=纸浆.lpSum([u[i]-u[j]+n*x[i,j]])0:
#打印str(i)+'''.'+str(j)+':'+str(纸浆值(x[i,j]))
查找_tour()
最后一个约束不是单个约束。您应该为满足该条件的每对索引i,j
添加一个约束:
for i in range(n-1):
for j in range(i+1, n):
m += pulp.lpSum([ u[i] - u[j] + n*x[i,j]]) <= n-1
你的代码有什么问题?@kindall我没有包括最后的约束,因为我不知道如何定义它。此外,对当前模型的求解给出了决策变量,如
x1_1
等于1.0
,这肯定是错误的。。。我不明白为什么会这样。@kindall更新了问题以澄清问题,并在当前代码中包含了不正确的结果。您将无法使用此方法解决更大的问题,因为由于子巡更消除约束,公式将变得非常大。您可以考虑另一种动态分离无效解决方案的方法。这里有一个Python示例:@mattmilten感谢您的建议,非常有用,因为我对TSP和LP还不熟悉,不知道什么方法更有效,特别是在现实生活中使用的方法。谢谢,我正在研究您的建议。到目前为止,当前代码看起来正常吗?缺少最后一个约束是不是因为x0_0
,x1_1
等被解决为具有1.0
的值而不是0.0
?@Nyxynyx最后一个约束对于问题来说是绝对基本的。没有它,问题甚至不是一个整数线性问题,而只是一个线性问题,因此它不能以任何有意义的方式表示TSP(请记住,TSP不能在多项式时间内求解,甚至不能在多项式时间内以“好的方式”近似,因此每个非整数线性规划问题都是如此)(总是可以在多项式时间内求解)不能表示它或任何一种良好的近似)当我试图同化最后一个约束的代码时,我将它们添加到当前代码并尝试求解它,但结果仍然很奇怪,变量x[I,j]的值为1.0
其中i==j
@Nyxynyx我相信您的代码中可能有错误。在原始模型中,如果我没有弄错的话,索引为0
的节点应该是“关闭循环”的节点。您可以看到,在目标函数中,您有从i=0
到n
的外部和。但是,其他约束从i=1
开始。在您的代码中,似乎您认为模型只是1
-索引,因此将约束更改为0
-索引…@Nyxynyx您应该尝试定义您的x
变量使用范围(n+1)中的i的:范围(n+1)中的j的,并修改所有约束以从1开始,因此即使最后一个约束也应该是范围(1,n)中的i的:范围(i+1,n+1)中的j的:…
。此外,没有u\u 0
变量,因此在定义u
向量时,您必须对此进行调整…只需添加一个伪en