在约束中对决策变量进行排序（引入新变量）

变量

约束

代码

用gurobi跑一跑，把xmk和ymk作为已知量。

import sys
from gurobipy import *
from decimal import Decimal
from math import *
import os
import random
import pandas as pd

from numpy import *

num_leg = 7
num_yard = 6
demand = 1
set_k = num_leg + num_yard #相当于r

x = [[1,0,0,1,0,1,0]] # xmk
y = [[0,0,0,0,1,0]] #ymk
cost = [60,50,40,35,55,45,75,80,80,80,80,80,80]  # c代表cost，cl和ck都在一个c里，只不过前面对应的是leg，后面对应的是yard

# Create a new model
hazmat = Model('hazmat')

# **********************
# ** Create variables **
# **********************

# zmk 表示m路线中经过的(i,j)和yard的集合
z = [0] * demand
for m in range(demand):
    z[m]=[0] * set_k
    for k in range(set_k):
        z[m][k] = hazmat.addVar(0.0, 1.0, 0.0, GRB.BINARY, 'z[%d][%d]' % (m,k))
        
# πmkr
π = [0] * demand
for m in range(demand):
    π[m] = [0] * set_k
    for k in range(set_k):
        π[m][k] =  [0] * set_k
        for r in range(set_k):
            π[m][k][r] = hazmat.addVar(0.0, 1.0, 0.0, GRB.BINARY, 'π[%d][%d][%d]' % (m,k,r))

# cmr
c = [0] * demand
for m in range(demand):
    c[m] = [0] * set_k
    for r in range(set_k):
        c[m][r] = hazmat.addVar(0.0, 100, 1.0, GRB.CONTINUOUS, 'c[%d][%d]' % (m,r))

hazmat.update()    

#************************************************************
# ** Add constraints **
#************************************************************
#   LinExpr.addTerms(coeffs, vars) 要一个系数乘上一个变量

# 对应约束(1), 对于leg：zmk=xmk
for m in range(demand):
    for k in range(num_leg):
        lhs = LinExpr()
        lhs.addTerms(1.0, z[m][k])
        hazmat.addConstr(lhs, GRB.EQUAL, x[m][k], 'c1(%d,%d)' %(m,k))

# 对应约束(2), 对于yard：zmk=xmk
for m in range(demand):
    for k in range(num_leg,num_leg+num_yard):
        lhs = LinExpr()
        lhs.addTerms(1.0, z[m][k])
        hazmat.addConstr(lhs, GRB.EQUAL, y[m][k-num_leg], 'c2(%d,%d)' %(m,k))

# 对应约束(3), πmkr<=zmk
for m in range(demand):
    for k in range(set_k):
        for r in range(set_k):
            lhs = LinExpr()
            lhs.addTerms(1.0, π[m][k][r])
            lhs.addTerms(-1.0, z[m][k])
            hazmat.addConstr(lhs, GRB.LESS_EQUAL, 0.0, 'c3(%d,%d,%d)' %(m,k,r))

# 对应约束(4), 累加r的πmkr<=1
for m in range(demand):
    for k in range(set_k):
        lhs = LinExpr()
        lhs.addTerms(-1.0, z[m][k])
        for r in range(set_k):
            lhs.addTerms(1.0, π[m][k][r])
        hazmat.addConstr(lhs, GRB.EQUAL, 0.0, 'c4(%d,%d)' %(m,k))

# 对应约束(5), 累加m和k的πmkr<=1
for m in range(demand):
    for r in range(set_k):
        lhs = LinExpr()
        for k in range(set_k):
            lhs.addTerms(1.0, π[m][k][r])
        hazmat.addConstr(lhs, GRB.LESS_EQUAL, 1.0, 'c5(%d,%d)' %(m,r))

# 对应约束(6), 累加m和k的πmkr<=1
for m in range(demand):
    lhs = LinExpr()
    for k in range(set_k):
        lhs.addTerms(-1.0, z[m][k])
        for r in range(set_k):
            lhs.addTerms(1.0, π[m][k][r])
    hazmat.addConstr(lhs, GRB.EQUAL, 0.0, 'c6(%d)' %(m))

# 对应约束(7), 累加r的πmkr=ymk
for m in range(demand):
    for r in range(set_k-1):
        lhs = LinExpr()
        for k in range(num_leg+num_yard):
            lhs.addTerms(1.0, π[m][k][r+1])
            lhs.addTerms(-1.0, π[m][k][r])
        hazmat.addConstr(lhs, GRB.LESS_EQUAL, 0.0, 'c7(%d,%d)' %(m,r))

# 对应约束(8), ck*πmkr>=cl*πmlr
for m in range(demand):
    for r in range(set_k-1):
        lhs = LinExpr()
        for k in range(num_leg+num_yard):
            lhs.addTerms(-cost[k], π[m][k][r])
        for l in range(num_leg+num_yard):
            lhs.addTerms(cost[l], π[m][l][r+1])
        hazmat.addConstr(lhs, GRB.LESS_EQUAL, 0.0, 'c8(%d,%d)' %(m,r))

# 对应约束(9), cmr=cl*πmlr
for m in range(demand):
    for r in range(set_k):
        lhs = LinExpr()
        lhs.addTerms(1.0, c[m][r])
        for k in range(num_leg+num_yard):
            lhs.addTerms(-cost[k], π[m][k][r])
        hazmat.addConstr(lhs, GRB.EQUAL, 0.0, 'c10(%d,%d)' %(m,r))

此代码对应着下图的路线，其中令x和y的值都是已知的，但在实际运行中，x和y是决策变量。

然后得到结果如下：

<gurobi.Var c[0][0] (value 80.0)>
<gurobi.Var c[0][1] (value 60.0)>
<gurobi.Var c[0][2] (value 45.0)>
<gurobi.Var c[0][3] (value 35.0)>
<gurobi.Var c[0][4] (value 0.0)>
<gurobi.Var c[0][5] (value 0.0)>
<gurobi.Var c[0][6] (value 0.0)>
<gurobi.Var c[0][7] (value 0.0)>

即完成了对路线中的c从大到小进行排序。