Infeasible model
Awaiting user inputHi,
I'm trying to solve an MIP model to optimality using Gurobi. The thing is that once I increase the power of one of my decision variables, for instance to 4, the model gets infeasible. I did not change the constraints, how can a model become infeasible by just making a change to its objective function? Here's my model, x, y, z, and y2, are the variables and the issue is with y2[j,k]:
Main = Model(name='main')
Main.Params.LogToConsole = 1
x = Main.addMVar((m, n), vtype = GRB.BINARY)
y = Main.addMVar((n, p), vtype = GRB.CONTINUOUS)
y2 = Main.addMVar((n, p), vtype = GRB.CONTINUOUS)
z = Main.addMVar((n), vtype = GRB.BINARY)
Main.setParam("NonConvex", 2)
obj = (quicksum(a[j]*z[j] for j in range (n))
+ quicksum(c[j,k]*(y2[j,k]**2) for j in range (n) for k in range (p))
+ quicksum(t[i,j]*x[i,j]* (D2[i]**4) for i in range (m) for j in range (n)))
Main.setObjective(obj, GRB.MINIMIZE)
for j in range(n):
for k in range(p):
Main.addGenConstrPow(y[j, k], y2[j, k], 2)
Main.addConstrs(quicksum(x[i, j] for j in range (n)) == 1 for i in range (m))
Main.addConstrs(quicksum(D[i]*x[i, j] for i in range (m)) ==
quicksum(y[j,k] for k in range (p)) for j in range (n))
Main.addConstrs(y[j,k] >= 0 for j in range (n) for k in range (p))
Main.addConstrs(y2[j,k] >= 0 for j in range (n) for k in range (p))
Main.addConstrs(quicksum(y[j,k] for k in range (p)) <= v[j]*z[j] for j in range (n))
Main.addConstrs(quicksum(D[i]*x[i,j] for i in range (m)) <= v[j]*z[j] for j in range (n))
Main.optimize()
-
Official comment
This post is more than two years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum, or try Gurobot, our chatbot interface offering instant, expert-level support. -
Hi Ensieh,
Can you extend your code so that it is a reproducible example? This will make it easier for someone from the community to lend help.
- Riley
0 -
Sure! Please see below.
n = 20
m = 20
p = 1
betha1 = 0.567 / 1000
a = np.full(n, 55)
scale = 4
v = np.ceil(np.array([1046., 1046., 627.6, 1046., 627.6, 1046., 1046., 627.6,
627.6, 627.6, 1046., 627.6, 627.6, 1046., 627.6, 627.6,
1046., 627.6, 1046., 1046.]))
df = pd.read_excel('distances.xlsx')
d_ij = df.values
d_ij = np.tile(d_ij, (scale, 1))
d_jk_l = [40, 440, 20, 0, 60, 170, 50, 30, 80, 340, 50, 390, 90, 90, 110, 70, 300, 50, 1390, 80]
d_jk = np.array(d_jk_l).reshape(-1, 1)
D = np.array([626, 241, 178, 151, 151, 78, 87, 62, 60, 48, 53, 39, 32, 35, 40, 20, 27, 30, 28, 37])
D = np.tile(D, scale)
D = D.astype('int64')c = np.round(np.multiply(betha1, d_jk), 5)
t = np.round(np.multiply(betha1, d_ij), 5)Main = Model(name='main')
Main.Params.LogToConsole = 0
Main.Params.MIPGap = 1e-6
Main.Params.OptimalityTol = 1e-6
Main.Params.IntFeasTol = 1e-6
Main.Params.Presolve = 2
Main.Params.ImproveStartGap = 1e-6
Main.Params.NumericFocus = 3
x = Main.addMVar((m, n), vtype = GRB.BINARY)
y = Main.addMVar((n, p), vtype = GRB.CONTINUOUS)
y2 = Main.addMVar((n, p), vtype = GRB.CONTINUOUS)
z = Main.addMVar((n), vtype = GRB.BINARY)
D2 = np.power(D, 4)
Main.setParam("NonConvex", 2)
obj = (quicksum(a[j]*z[j] for j in range (n))
+ quicksum(c[j,k]*(y2[j,k]) for j in range (n) for k in range (p))
+ quicksum(t[i,j]*x[i,j]* (D2[i]) for i in range (m) for j in range (n)))
Main.setObjective(obj, GRB.MINIMIZE)
for j in range(n):
for k in range(p):
Main.addGenConstrPow(y[j, k], y2[j, k], 4)
Main.addConstrs(quicksum(x[i, j] for j in range (n)) == 1 for i in range (m))
Main.addConstrs(quicksum(D[i]*x[i, j] for i in range (m)) ==
quicksum(y[j,k] for k in range (p)) for j in range (n))
Main.addConstrs(y[j,k] >= 0 for j in range (n) for k in range (p))
Main.addConstrs(y2[j,k] >= 0 for j in range (n) for k in range (p))
Main.addConstrs(quicksum(y[j,k] for k in range (p)) <= v[j]*z[j] for j in range (n))
Main.addConstrs(quicksum(D[i]*x[i,j] for i in range (m)) <= v[j]*z[j] for j in range (n))
Main.optimize()
print(Main.objval)0 -
Hi Ensieh,
df = pd.read_excel('distances.xlsx')
Your code is not reproducible.
- Riley
0
Post is closed for comments.
Comments
4 comments