Warning of constraint limit exceed fixing

Ongoing

sujana islam

• February 08, 2023 15:15

The following information for my question is provided below:
I am trying to solve

here M=26=i
Btotal=5*1.5e6

Gain_final_V2V=giv2i-bs(of the pic)=
[2597.5727990651158, 3156.978351073531, 5305.489326799453, 3488.9747383001554, 3556.3703346242582, 4209.642834042834, 3216.2253501689943, 3616.6713767731144, 3721.2489122703337, 4815.242406071199, 7523.679665260634, 5148.824520601632, 4781.0405235624285, 4825.401541460312, 5293.640310589527, 7880.490491309103, 5144.549521524911, 4058.1748037853367, 4029.064234977265, 3670.7731966185406, 4933.650603653963, 4301.815468714573, 4629.004866144499, 7406.363983954186, 4736.348196770436, 4446.8020800583445]

and Gain_final_V2V=gjv2v-bs(of the pic)
=[3411.7021800330017, 4383.893622532522, 3143.920480566424, 3152.1430240894356, 2596.579324386922, 3044.244764660792, 3524.4518965433053, 2510.4417141729073, 2863.8285361945786, 5714.688173071876, 3152.3522620213735, 2668.9943712772624, 2173.361880068201, 3192.161712881365, 4462.116286156692, 1699.3377755789538, 2835.485261732316, 2012.5175075008174, 3480.850305355655, 3853.166079972098]

for O in range(26):

  from gurobipy import Model, GRB

  # Create a new model

  m = Model()

  # Define decision variables p[i, j] and pjv2v

  p = {}

  pjv2v = []

  piv2i=[]

  for j in range(20):

      p[1, j+1] = m.addVar(lb=0, vtype=GRB.BINARY, name="p[1,{}]".format(j+1))

      pjv2v.append(m.addVar(lb=-100, ub=23, name="pjv2v[{}]".format(j+1)))

  for i in range(26):

      piv2i.append(m.addVar(lb=-100, ub=23, name="pjv2v[{}]".format(i+1)))

  # Update the model to include the variables

  m.update()

  # Define the constraint

  con = sum(p[1, j+1] for j in range(20)) <= 3 ##eqn_2 of base ppr

  # Add the constraint to the model

  m.addConstr(con)

  noise=[]

  rho_n=-(174) #in dbm/hz

  Btotal=5*1.5e6

  noise=(Btotal*rho_n)/26

    

  # Define the objective function

  down = sum(p[1, j+1] * Gain_final_V2V[j] * pjv2v[j] for j in range(20))+noise

  

  

  interference=sum(p[1, j+1] * Gain_final_V2V[j] * pjv2v[j] for j in range(20))

  up=piv2i[O]*Gain_final_V2I[O]

  eqn_11f=m.addConstrs(( piv2i[i] <=23 for i in range(26)), name='eqnf')

   

  eqn_11g= m.addConstrs(( pjv2v[j] <=23 for j in range(20)), name='eqng')

  

  auxiliary_variable=m.addVar(lb=-GRB.INFINITY, ub = -1, vtype=GRB.CONTINUOUS, name="Auxiliary_Variable")

  m.addConstr(auxiliary_variable==down)




  SINR=m.addVar(lb=-GRB.INFINITY,ub=GRB.INFINITY,vtype=GRB.CONTINUOUS, name="SINR")

  abc=SINR*auxiliary_variable

  m.addConstr(abc==up)

  m.params.NumericFocus = 2

  #m.setObjective(SINR,GRB.MINIMIZE)

  w=m.addVar(lb=0,ub=GRB.INFINITY, vtype=GRB.CONTINUOUS, name="1+SINR")

  m.addConstr(w==1+SINR)

  z=m.addVar(lb=0,ub=GRB.INFINITY, vtype=GRB.CONTINUOUS, name="log(1+SINR)")

  gc = m.addGenConstrLogA(w, z, 2)

  m.setObjective(z,GRB.MAXIMIZE)

   

  m.setParam("NonConvex",2)

  m.optimize()

but here whenever I run the code the error that pops up is:
how to solve this
kindly help :)

Related to

• General Constraints

0

• 

  sujana islam

  • February 09, 2023 08:55

  

  

  

  also i am having another problem.
  

  here N=20
  M=26
  and z=3
  
  meaning the eqn:
  

  in this eqn the part of denumator sometimes should take 1 ,2 and 3 as summing p[i,j] like the previous pic
  
  but in my following code it is always taking 3 
  how to fix this

  26 M should all together handle 20 N
  but in my code it is handling 26*3 N
  because it is always taking 3
  but I want to do that it should take 1 ,2 or 3 sometimes as the sum of 1st snap and alltogether handle 20Ns
  how i can fix this
  kindly help
  
  best regards
  Sujana

  0
• 

  Riley Clement

  Gurobi Staff

  • February 09, 2023 12:17
  • Edited

  Hi Sujana,

  In your code you use the `Model.addGenConstrLogA` function.  The documentation for this function is here and describes four attributes which control the piecewise linearization that Gurobi performs: FuncPieces, FuncPieceError, FuncPiecesLength, and FuncPieceRatio.  Your warning message is suggesting that you make adjustments to these attributes to try and reduce the size of the constraint violations in the solution found.  Finding the right values to satisfactorily decrease the violations while achieving a satisfactory solve time may be a process of trial and error.

  Respectfully, I do not understand your second question, but perhaps someone else in the community will.  You may find inspiration in your colleague's related question.

  Best of luck,
  Riley

  0
• 

  sujana islam

  • February 09, 2023 15:05

  

  

  

  thanks for respoding Riley
  my ques is :
  
  

  here M=26

  N=20

  z=3

  and  

  

  now my objective is to:

  

  now in my code i am unable to include

  this portion and even tho my code works it is always taking the p[1,j] values altho it should take p[i,j] after each iteration where i=26,j=20
  
  and my following code is:

  Gain_final_V2V=giv2i-bs(of the pic)=
  [2597.5727990651158, 3156.978351073531, 5305.489326799453, 3488.9747383001554, 3556.37Gain_final_V2V=gjv2v-bs(of the pic)
  Gain_final_V2V=gjv2v-bs(of the pic)
  =[3411.7021800330017, 4383.893622532522, 3143.920480566424, 3152.1430240894356, 2596.579324386922, 3044.244764660792, 3524.4518965433053, 2510.4417141729073, 2863.8285361945786, 5714.688173071876, 3152.3522620213735, 2668.9943712772624, 2173.361880068201, 3192.161712881365, 4462.116286156692, 1699.3377755789538, 2835.485261732316, 2012.5175075008174, 3480.850305355655, 3853.166079972098]03346242582, 4209.642834042834, 3216.2253501689943, 3616.6713767731144, 3721.2489122703337, 4815.242406071199, 7523.679665260634, 5148.824520601632, 4781.0405235624285, 4825.401541460312, 5293.640310589527, 7880.490491309103, 5144.549521524911, 4058.1748037853367, 4029.064234977265, 3670.7731966185406, 4933.650603653963, 4301.815468714573, 4629.004866144499, 7406.363983954186, 4736.348196770436, 4446.8020800583445]
  
  

  for O in range(26):
  
    from gurobipy import Model, GRB
  
    # Create a new model
  
    m = Model()
  
    # Define decision variables p[i, j] and pjv2v
  
    p = {}
  
    pjv2v = []
  
    piv2i=[]
  
    for j in range(20): #@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
  
        p[O+1, j+1] = m.addVar(lb=0, vtype=GRB.BINARY, name="p[1,{}]".format(j+1))
  
        pjv2v.append(m.addVar(lb=-100, ub=23, name="pjv2v[{}]".format(j+1)))
  
    for i in range(26):
  
        piv2i.append(m.addVar(lb=-100, ub=23, name="pjv2v[{}]".format(i+1)))
  
    # Update the model to include the variables
  
    m.update()
  
    # Define the constraint
  
    con = sum(p[O+1, j+1] for j in range(20)) <= 3 ##eqn_2 of base ppr #@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
  
    print(con)
  
    
  
    
  
    noise=[]
  
    rho_n=-(174) #in dbm/hz
  
    Btotal=5*1.5e6
  
    noise=(Btotal*rho_n)/26
  
      
  
    # Define the objective function
  
    down = sum(p[O+1, j+1] * Gain_final_V2V[j] * pjv2v[j] for j in range(20))+noise #@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
  
    print(down)
  
    
  
    up=piv2i[O]*Gain_final_V2I[O]
  
    eqn_11f=m.addConstrs(( piv2i[i] <=23 for i in range(26)), name='eqnf')
  
     
  
    eqn_11g= m.addConstrs(( pjv2v[j] <=23 for j in range(20)), name='eqng')  #@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
  
    
  
    auxiliary_variable=m.addVar(lb=-GRB.INFINITY, ub = -1, vtype=GRB.CONTINUOUS, name="Auxiliary_Variable")
  
    m.addConstr(auxiliary_variable==down)
  
  
  
  
    SINR=m.addVar(lb=-GRB.INFINITY,ub=GRB.INFINITY,vtype=GRB.CONTINUOUS, name="SINR")
  
    abc=SINR*auxiliary_variable
  
    m.addConstr(abc==up)
  
    m.params.NumericFocus = 2
  
    #m.setObjective(SINR,GRB.MAXIMIZE)
  
    w=m.addVar(lb=0,ub=GRB.INFINITY, vtype=GRB.CONTINUOUS, name="1+SINR")
  
    m.addConstr(w==1+SINR)
  
    LOG2_=m.addVar(lb=0,ub=GRB.INFINITY, vtype=GRB.CONTINUOUS, name="log2(1+SINR)")
  
    gc = m.addGenConstrLogA(w, LOG2_, 2)
  
    aaa=(5*1.5e9)/26
  
    m.setObjective((aaa*LOG2_),GRB.MAXIMIZE)
  
     
  
    m.setParam("NonConvex",2)
  
    m.optimize()

  
  now whenever i print con then it always gives parameter with p[1,*] means O is not changing

  kindly tell me how i can incorporate the eqn i am unable to and finally do the objective line
  
  badly in need of your expertise on board
  best regards,

  sujana

  0
• 

  Riley Clement

  Gurobi Staff

  • February 10, 2023 01:12

  Hi Sujana,
  
  Your model may be correct but you are not giving your variables unique names and so your "con" constraints (which are different) look the same when printing.
  
  Replace

  name="p[1,{}]".format(j+1)

  with

  name="p[{},{}]".format(O+1,j+1)

  to fix the issue.

   

  - Riley

   

  0
• 

  sujana islam

  • February 10, 2023 02:41

  

  

  

  thank you riley but i am still unable to incorporate 
  
  inside the loop
  and when i try that 

  
   constr = sum(p[i+1, 20] for i in range(26) if (i+1, 20) in p) <= 1 
   m.addConstr(constr)
   print(constr)

  inside that for O in range(26):
  it always provides:
  
  how to fix this
  i want p[1,20]+p[2,20]+p[3,20]................+p[26,20]<=1
  
  also i am unsure of whether my code satisties

  seeking your help in both the aspects

  
  Best regards,
  sujana

  0
• 

  Riley Clement

  Gurobi Staff

  • February 10, 2023 04:43
  • Edited

  Hi Sujana,

  You are looping and increasing the value of O as you loop. You are creating a fresh dictionary p each time in the loop and add items whose first index is O+1. As such your dictionary will only contain variables whose first indice is identical. Are you aware that your code is creating 26 models? It is not clear to me whether you are intending this.

  I am also unsure of whether your code satisfies the equation.

  - Riley

  0
• 

  sujana islam

  • February 10, 2023 08:37

  

  

  

  thank you for your response riley
  yes I m creating 26 models to get 26 values of (Btotal/M)log2(1+sinr) cause i am unable to sum up as my equation is stated

  

  hence afterwards getting 26 values I am manually summing up the best objective
  if you know how i can sum up all of them at once please let me know
  
  without using the loop how this can be done please let me know
  
  

  provided constraints are:
  

  
  and infos of them is stated above comment
  
  with regards,

  sujana

   

  0
• 

  Riley Clement

  Gurobi Staff

  • February 10, 2023 10:38

  Hi Sujana,

  I would check with your teacher if this is the best approach. Or perhaps Tanbir Mahatab Emon can help?

  - Riley

  0
• 

  sujana islam

  • February 10, 2023 17:39

  

  

  

  hi riley,

  you were right

  now i could implement the problem by 1 model
  where i have taken i=5,j=5
  afterward i=5,j=10 and i=5,j=15
  
  all of them shows the same result as below:
  why is this happening?
  even though i am changing j manually to record no matter what is j it is giving these 2 values only
  
  my following code right now is:
  
  Gain_final_V2I=[-1556.969928074958, 115088.91771640752, 177694.47911317323, 130573.06734283901, -2659.099484648859]
  
  
  

  when j=5:

  Gain_final_V2V=[-10440.614079695104, 10.739604231320882, -16601.94581425221, -20400.772858467815, -24726.782460859096]

  from gurobipy import Model, GRB
  
  m = Model()
  
  # Define decision variables p[i, j] and pjv2v
  
  p = {}
  
  pjv2v = []
  
  piv2i=[]
  
  for O in range(5):
  
    for j in range(5): 
  
        p[O+1, j+1] = m.addVar(lb=0, vtype=GRB.BINARY, name="p[{},{}]".format(O+1,j+1))
  
        
  
  for i in range(5):
  
     piv2i.append(m.addVar(lb=0, ub=23, name="piv2i[{}]".format(i+1)))
  
  for j in range(5):
  
    pjv2v.append(m.addVar(lb=0, ub=23, name="pjv2v[{}]".format(j+1)))
  
  m.update()
  
  print(p)
  
  print(piv2i)
  
  print(pjv2v)
  eqn_1=m.addConstrs(sum(p[i+1,j+1] for i in range(5))<=1 for j in range(5))
  
  eqn_2=m.addConstrs(sum(p[i+1,j+1] for j in range(5))<=3 for i in range(5))
  
  rho_n=-(174) #in dbm/hz
  
  Btotal=5*1.5e6
  
  noise=(Btotal*rho_n)/5
  m.setObjective(sum(final),GRB.MAXIMIZE)
  
  m.setParam("NonConvex",2)
  
  m.optimize()
  
  

  when j=15:
  Gain_final_V2V=[-20883.423226544437, -8525.550572680211, -14971.827485316973, -43467.82648846318, -21180.286507383677, -14245.638762119452, -43749.53738731502, -13490.670399685234, -10564.329882674141, -16388.62883316595, -9434.96993886823, -3342.047080598144, -11025.490544368218, -19638.213224785184, -18993.831432557614]
  
  

  from gurobipy import Model, GRB
  
  m = Model()
  
  # Define decision variables p[i, j] and pjv2v
  
  p = {}
  
  pjv2v = []
  
  piv2i=[]
  
  for O in range(5):
  
    for j in range(15): 
  
        p[O+1, j+1] = m.addVar(lb=0, vtype=GRB.BINARY, name="p[{},{}]".format(O+1,j+1))
  
        
  
  for i in range(5):
  
     piv2i.append(m.addVar(lb=0, ub=23, name="piv2i[{}]".format(i+1)))
  
  for j in range(15):
  
    pjv2v.append(m.addVar(lb=0, ub=23, name="pjv2v[{}]".format(j+1)))
  
  m.update()

  eqn_1=m.addConstrs(sum(p[i+1,j+1] for i in range(5))<=1 for j in range(15))
  
  eqn_2=m.addConstrs(sum(p[i+1,j+1] for j in range(15))<=3 for i in range(5))

  rho_n=-(174) #in dbm/hz
  
  Btotal=5*1.5e6
  
  noise=(Btotal*rho_n)/5
  
  

  down = [0] * 5
  
  up=[0]*5
  
  auxiliary_variable=[0]*5
  
  SINR=[0]*5
  
  abc=[0]*5
  
  w=[0]*5
  
  LOG2_=[0]*5
  
  gc=[0]*5
  
  aaa=[0]*5
  
  final=[0]*5
  
  for k in range(5):
  
    down[k]= sum(p[k+1,j+1]*pjv2v[j]*Gain_final_V2V[j] for j in range(15))+noise
  
    up[k]=piv2i[k]*Gain_final_V2I[k]
  
    auxiliary_variable[k]=m.addVar(lb=-GRB.INFINITY ,ub =-1, vtype=GRB.CONTINUOUS, name="Auxiliary_Variable")
  
    m.addConstr(auxiliary_variable[k]==down[k])
  
  
  
  
    SINR[k]=m.addVar(lb=0,ub=GRB.INFINITY,vtype=GRB.CONTINUOUS, name="SINR")
  
    abc[k]=SINR[k]*auxiliary_variable[k]
  
    m.addConstr(abc[k]==up[k])
  
    m.params.NumericFocus = 2
  
    w[k]=m.addVar(lb=1,ub=GRB.INFINITY, vtype=GRB.CONTINUOUS, name="1+SINR")
  
    m.addConstr(w[k]==1+SINR[k])
  
    LOG2_[k]=m.addVar(lb=-GRB.INFINITY,ub=GRB.INFINITY, vtype=GRB.CONTINUOUS, name="log2(1+SINR)")
  
    gc[k] = m.addGenConstrLogA(w[k], LOG2_[k], 2)
  
    aaa[k]=(5*1.5e6)/5
  
    
  
    final[k]=aaa[k]*LOG2_[k]
  
  

  m.setObjective(sum(final),GRB.MAXIMIZE)
  
  m.setParam("NonConvex",2)
  
  m.optimize()

  but these 2 give the same result but there should be a change for the change in j 
  why no matter whatever j i am giving it is providing either 803 or 4.33
  
  kindly help
  where am i going wrong
  
  best regards
  sujana

   

  0

Please sign in to leave a comment.