Professional Documents
Culture Documents
Jain, V.,& Grossmann, I. E. (2001) - Algorithms For Hybrid MILP-CP Models For A Class of Optimization Problems
Jain, V.,& Grossmann, I. E. (2001) - Algorithms For Hybrid MILP-CP Models For A Class of Optimization Problems
This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
approval, unless otherwise noted. For more information, contact permissions@informs.org.
The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
support of claims made of that product, publication, or service.
© 2001 INFORMS
INFORMS is the largest professional society in the world for professionals in the fields of operations research, management
science, and analytics.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
Algorithms for Hybrid MILP/CP Models
for a Class of Optimization Problems
Vipul Jain • Ignacio E. Grossmann
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
Department of Chemical Engineering, Carnegie Mellon University, Pittsburgh, Pennsylvania, 15213, USA
Vipul_Jain@i2.com • grossmann@cmu.edu
T he goal of this paper is to develop models and methods that use complementary
strengths of Mixed Integer Linear Programming (MILP) and Constraint Programming
(CP) techniques to solve problems that are otherwise intractable if solved using either of
the two methods. The class of problems considered in this paper have the characteristic
that only a subset of the binary variables have non-zero objective function coefficients if
modeled as an MILP. This class of problems is formulated as a hybrid MILP/CP model
that involves some of the MILP constraints, a reduced set of the CP constraints, and equiv-
alence relations between the MILP and the CP variables. An MILP/CP based decomposi-
tion method and an LP/CP-based branch-and-bound algorithm are proposed to solve these
hybrid models. Both these algorithms rely on the same relaxed MILP and feasibility CP
problems. An application example is considered in which the least-cost schedule has to be
derived for processing a set of orders with release and due dates using a set of dissim-
ilar parallel machines. It is shown that this problem can be modeled as an MILP, a CP,
a combined MILP-CP OPL model (Van Hentenryck 1999), and a hybrid MILP/CP model.
The computational performance of these models for several sets shows that the hybrid
MILP/CP model can achieve two to three orders of magnitude reduction in CPU time.
(Integer Programming; Benders-Decomposition; Constraint Programming; MILP/CP Hybrid
Algorithms; Parallel Machine Scheduling )
We formulate this class of problems as hybrid or the LP subproblem is infeasible. Branching is per-
MILP/CP models that involve some of the MILP con- formed whenever the solution obtained by solving
straints, a reduced set of the CP constraints, and the LP relaxation does not satisfy all the constraints
the equivalence relations between the MILP and the in the original problem. If the relaxed solution sat-
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
CP variables. We then propose decomposition and isfies all the constraints in the original problem and
branch-and-bound algorithms to solve these hybrid is better than the best feasible solution found so far,
models. Both of these algorithms rely on relaxed then the best feasible solution is updated. The search
MILP and feasibility CP problems. The aim of these terminates when it is proved that no better solution
methods is to combine the strength of MILP for prov- exists than the best feasible solution found.
ing optimality by using the LP relaxations, and the The effectiveness of MILP methods depends on
power of CP for finding feasible solutions by using the size of the linear-programming subproblems, and
the specialized propagation algorithms. As an exam- more importantly, on the gap between the objective
ple, we consider a scheduling problem that involves for the best feasible solution (optimum) and the objec-
finding a least-cost schedule to process a set of orders tive function value obtained from the initial LP sub-
using dissimilar parallel machines subject to release problem. There are a number of more sophisticated
and due-date constraints. It is shown that this prob- algorithms that focus on these aspects and use differ-
lem can be modeled as an MILP, CP, or a combined ent ways of generating LP subproblems, like Branch
MILP-CP OPL model (Van Hentenryck 1999). We then and Cut (Padberg and Rinaldi 1991, Balas et al. 1996),
investigate the computational performance of these and Branch and Price (Barnhart et al. 1998). These
alternative models for a number of data sets and algorithms make use of valid inequalities to improve
highlight the advantages and disadvantages of these the performance of the solution algorithms.
approaches. This problem is then modeled as a hybrid CP, on the other hand, uses constraint propaga-
MILP/CP model and is solved using the proposed tion as the inference engine. At each node, con-
decomposition algorithm. Computational results are straint propagation is used to reduce the domains of
presented and finally some conclusions are drawn. all the variables. The domain of a variable can be
continuous, discrete, boolean etc. Constraint propa-
gation can result in empty domains in which case
2. Background a node is fathomed. Branching is performed when-
MILP-based methods were developed over the last ever the domain of a variable consists of more than
four decades by the Operations Research commu- one element (discrete and boolean domains) or the
nity (Nemhauser and Wolsey 1988). CP-based meth- bounds are not within a certain tolerance (continuous
ods, on the other hand, are the result of research domains). CP was originally developed to solve fea-
by the Artificial Intelligence community in the area sibility problems. It has now been extended to solve
of Logic Programming and Constraint Satisfaction optimization problems. This is achieved by solving a
(Colmerauer 1985, Van Hentenryck 1989, Tsang 1993). feasibility problem in which the objective function of
Both these frameworks rely on branching to explore the problem is rewritten as a constraint that forces
the search space. The primary difference lies in the it to be equal to a new variable. The domain of this
way inference is performed at each node. Linear new variable gives upper and lower bounds on the
Programming (LP)-based branch-and-bound meth- objective-function values. Whenever a feasible solu-
ods for MILP involve solving LP subproblems that tion is obtained during the search, additional con-
are generated by dropping some of the constraints straints that restrict the objective-function values are
(integrality constraints) to obtain bounds on the imposed throughout the search tree. The search ter-
objective-function values and to prove that a set of minates when all the nodes have been fathomed. The
constraints is inconsistent. A node is fathomed either effectiveness of the CP methods depends primarily on
when the objective-function value of the LP relaxation the constraint-propagation algorithms that are used to
is worse than the best integer solution obtained so far, reduce the domain of a variable.
MILP has the advantage that the effect of all the (Proll and Smith 1998), the progressive party prob-
constraints is evaluated simultaneously and it has a lem (Smith et al. 1997), and the change problem
“Global Perspective” on all the constraints (Rodosek (Heipcke 1999a). Properties of a number of different
et al. 1999). CP, on the other hand, evaluates the effect
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
vis-a-vis pure MILP and CP approaches was signif- approaches. For example, probing and integer pre-
icantly better. Bockmayr and Kasper (1998) did an processing in MILP is in some ways similar to con-
interesting analysis of CP and MILP approaches, and straint propagation. Chandru and Hooker (1999) give
presented a unifying framework, Branch and Infer, that an interesting operations-research perspective on con-
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
can be used to develop various integration strate- sistency methods and logical inference. Also, Hooker
gies. They divide constraints for both MILP and (2000) deals with the subject of MILP and CP integra-
CP into two different categories, primitive and non- tion in detail.
primitive. Primitive constraints are those for which
there exists a polynomial-time solution algorithm, and
non-primitive constraints are those for which this is 4. Theory
not true. The interesting aspect about this classifica- The algorithms proposed in this section have been
tion is that some of the primitive constraints in CP motivated by the work of Bockmayr and Kasper
are non-primitive in MILP and vice versa. They also (1998). These algorithms are based on the premise that
discussed how non-primitive constraints can be used combinatorial problems may sometimes have some
to infer primitive constraints and the use of symbolic characteristics that are better suited for MILP and oth-
constraints for MILPs. Raman and Grossmann (1993, ers that are better handled by CP. For these problems,
1994) earlier modeled discrete/continuous optimiza- pure MILP- and pure CP-based approaches may not
tion problems with disjunctions and symbolic con- perform well. As discussed earlier, most of the prior
straints in the form of logic propositions. This model, work on integrating the two approaches use at least
which they denoted as a Generalized Disjunctive Pro- one of the models in complete form (usually CP). In
gram (GDP), can be converted all or in part into the algorithms that we present the problem is solved
an MILP. They presented the idea of w-MIP repre- using relaxed MILP and CP feasibility models.
sentability, which is similar to the idea of primitive Consider a problem, which when modeled as an
constraints. They showed that it is computationally MILP, has the following structure,
efficient to transform w-MIP representable disjunc-
tions into linear constraints, and proposed a hybrid (M1): min c T x (1)
branch-and-bound algorithm that handles the non st Ax + By + Cv ≤ a (2)
w-MIP representable disjunctions directly.
From the work that has been performed, it is A x + B y + C v ≤ a (3)
not clear whether a general integration strategy will n m
x ∈ 0 1 y ∈ 0 1 v ∈ p
(4)
always perform better than either a CP or an MILP
approach by itself. This is especially true for the This is an optimization problem that has both con-
cases where one of these methods is a very good tinuous (v) and binary (x and y) variables, and
tool to solve the problem at hand. However, it is only some of the binary variables (x) have non-zero
usually possible to enhance the performance of one objective-function coefficients. The constraint set can
approach by borrowing some ideas from the other. be divided into two subsets. The first set of con-
For example, Raman and Grossmann (1993) used straints (2) models some aspects of the problem that
logic cuts that were written as logic propositions to can be represented efficiently in the MILP framework
improve the performance of MILP models. Ideas on (e.g. assignment constraints) and has a significant
edge-finding (Carlier and Pinson 1989, Applegate and impact on the LP relaxation. The second set of con-
Cook 1991) that were used for guiding the search in straints (3), on the other hand, is assumed not to affect
MILPs to solve jobshop problems, were exploited by the LP relaxation significantly, and is sometimes large
Caseau and Laburthe (1994, 1996) and Le Pape (1994) in number because of the limited expressive power of
to develop efficient inference engines for scheduling MILP methods.
algorithms in CP. Furthermore, there are a number of The same problem can also be modeled as a
similarities in some of the underlying ideas of both CP. Note that more constructs are available in the
CP framework to model the problem (e.g. logical The hybrid model involves MILP constraints, CP
constraints, disjunctions, all-different operator etc; constraints, and equivalence relations. The objective
Marriot and Stuckey 1998). For this reason, the MILP function in the hybrid model (M3) is the same as in
and CP models of the same problem may have dif- the MILP model (M1). The constraints for this prob-
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
ferent variable definitions and constraint structures. lem include MILP constraints (2), equivalence rela-
However, an equivalence can be established between tions that relate MILP variables x to CP variables x̄,
the constraints and a complete labeling of variables and a reduced set of CP constraints that are derived
can be derived in one framework from the values of from the CP constraint set (6) by assuming that the
variables in the other one. Let us assume that the set of CP variables x̄ is fixed.
equivalent CP model is
and (13) of the model. If the partial solution from the models and play a very important role in the success
MILP can be extended, then the full-space solution of these methods. The entire procedure is repeated
obtained will also have the same value of the objec- until the solution obtained using the relaxed MILP
tive function. In this paper we present two methods to model can be extended to a full feasible solution, or
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
search the solution space and obtain the optimal solu- the relaxed MILP problem becomes infeasible.
tion. Both of these ideas are essentially the same and Assuming that the general “no good” cuts (14)
use the same relaxed MILP and CP feasibility models. are used we can establish the following convergence
However, the difference lies in the order in which the proof for the MILP/CP-based decomposition method.
CP subproblems are solved.
Theorem 1. If the MILP/CP decomposition method is
applied to solve problem (M3) with the cuts in (14), then
4.1. MILP/CP-Based Decomposition Method
the method converges to the optimal solution or proves
This algorithm has some similarities to Generalized
infeasibility in a finite number of iterations.
Benders Decomposition (Geoffrion 1972). The algo-
rithm is summarized in Figure 1. In this method a Proof. The problem (M3) is not unbounded
relaxed MILP model of the problem, with (8) as the because the domain of the variable x is bounded. So
objective and (9 and 12) as constraints, is solved to the problem can either be fesible or infeasible. Let us
optimality. Note that integrality constraints on the consider each of the cases individually.
variable y can be dropped if that makes the relaxed Case 1: Problem (M3) is feasible
MILP problem easier to solve. If there is no solution, The relaxed MILP master problem with cuts in (14)
then the original problem is infeasible. Otherwise, val- after iteration K is given by
ues of x are used to determine values of the equiva-
(RMK ) min z = c T x
lent CP variables x̄. A CP feasibility problem then tries
to extend this partial solution to a complete solution st Ax +By +Cv ≤ a
(x̄ ȳ v̄) that satisfies the constraints (11) and (13). If
xi − xi ≤ B k −1 ∀k ∈ 12K −1
there exists a feasible solution, then this solution is i∈T k i∈F k
the optimal solution of the problem and the search
T k = i
xik = 1F k = i
xik = 0B k =
T k
is terminated. Otherwise the causes for infeasibility
are inferred as cuts and added to the relaxed MILP ∀k ∈ 12K −1
model of the problem. These cuts could be general
x ∈ 01n y ∈ 01m v ∈ p
“no good” cuts (Hooker et al. 1999). In the context of
the model (M3) these cuts have the following form
for iteration k, Let xK be the optimal solution for this relaxed
MILP master problem. The corresponding CP feasi-
xi − xi ≤ B k − 1T k = i
xik = 1 (14)
bility subproblem is given by,
i∈T k i∈F k
Here, xk = x1k x2k represents the optimal values st Ḡx̄K ȳ v̄ ≤ 0
of x in iteration k. These general “no good” cuts may ȳ v̄ ∈
be rather weak. For this reason, whenever possible
stronger cuts (Qk x ≤ q k ) that exploit the special struc- where (xK ⇔ x̄K ). Note that for all iterations k < K,
ture of the problem should be used. The problem- the relaxed MILP master problems (RMk ) are feasible
specific cuts should not only cut off the current partial and the (CPk ) subproblems are infeasible. If either of
solution, but also eliminate partial solutions with sim- the two conditions are not satisfied the algorithm ter-
ilar characteristics. Cuts are a means of communi- minates before reaching iteration K. Since the master
cation between the relaxed MILP and CP feasibility problem for the first iteration is a relaxation of the
min c T x
✛
st Ax + By + Cv ≤ a
Qk x ≤ q k ∀k ∈
1 2 K − 1
x ∈
0 1n y ∈
0 1m v ∈ p
Feasible
❄
Optimal Solution Found
original problem, the MILP problem (RMK ) is also a 2n elements) and is non empty for problem (M3), it
relaxation of the problem (M3). This because the “no implies that the algorithm will reach a step where the
good” cuts that have been added until iteration K do (CPk ) subproblem is feasible. Assume that at iteration
not exclude any feasible solution from the original K, the subproblem (CPK ) yields a feasible solution
problem (M3). Furthermore, the domain of successive (x̄K ȳ v̄). This solution is optimal for problem (M3)
relaxation is strictly smaller than the previous one. as it belongs to the domain of (M3) and the objective-
Since the domain of x is finite (in the worst case has function value for this solution equals the optimum
the LP subproblems belonging to set P by updating partial solution will then involve both binary and con-
the set (Q q). Go to step 2. tinuous variables and the CP problem will extend this
4. Branch on a Variable. Of all the variables that partial solution in the full space. However, in such a
have been assigned non-integral values, select one case the biggest challenge is to derive effective cuts
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
according to some prespecified branching variable selec- that exclude partial solutions that are feasible for the
tion rule (Nemhauser and Wolsey 1988). Let us denote master problem but can not be extended in the full
this integer variable by zp . Generate two LP subprob- space.
lems p1 and p2 and add them to set P. The subproblem
p1 is generated by specifying the floor of the optimal
value of zp as the upper bound for zp , and the sub-
5. Scheduling Problem
We consider a specific scheduling problem that falls
problem p2 is generated by specifying the ceiling of
into the class of problems considered in this paper
the optimal value of zp as the lower bound of zp . Go
and that is similar to the one considered by Hooker
to step 2.
et al. (1999). This scheduling problem involves find-
5. Termination. If UB = , then the problem does
ing a least-cost schedule to process a set of orders I
not have a feasible solution or it is unbounded. Other-
using a set of dissimilar parallel machines M. Pro-
wise, the optimal solution corresponds to the current
cessing of an order i ∈ I can only begin after the
value UB.
release date ri and must be completed at the latest by
For the proposed decomposition and B&B meth-
the due date di . Order i can be processed on any of
ods to be successful, it is of course very important to
the machines. The processing cost and the processing
choose suitable relaxed MILP and CP feasibility mod-
time of order i ∈ I on machine m ∈ M are Cim and pim ,
els. Furthermore, rather than using the cuts in (14)
respectively. In this section, we consider three alter-
inferring strong cuts, Qx ≤ q, as causes for infeasibil-
native strategies to model and solve this particular
ity of CP feasibility models can significantly reduce
problem.
the number of problems to be solved. These cuts,
however, are non-trivial and should be derived when-
5.1. MILP Model
ever possible for each class of problems at hand. It is
The scheduling problem described above can be mod-
also worth emphasizing that the proposed algorithms
eled as an MILP. The main decisions involved in
do not impose any restriction on the structure of the
this scheduling problem are assignment of orders on
equivalence relation in (10). It can even be procedu-
machines, sequence of orders on each machine, and
ral. It should be noted that hybrid models similar to
start time for all the orders. The binary variable xim is
the ones presented in this section can also be solved
an assignment variable, and it is equal to one when
using OPL (Van Hentenryck 1999). In OPL, the solu-
order i is assigned to machine m. Binary variable yii
tion algorithm for the hybrid model solves an LP sub-
is the sequencing variable, and it is equal to one when
problem involving all the linear constraints, as well as
both i and i are assigned to the same machine and
a constraint-propagation subproblem involving all the
order i is processed after order i. The continuous vari-
constraints at every node of the search tree. Usually,
able tsi denotes the start time of order i. Using these
all the original CP constraints (6) are needed in the
variables the MILP model for this problem can be
hybrid MILP-CP OPL model. Furthermore, the equiv-
written as
alence relations (10) must be written in closed form
as equations, inequalities, or symbolic relations. min Cim xim (15)
Even though the algorithms presented in this i∈I m∈M
section are limited to the MILP models that have only s.t. tsi ≥ ri ∀i ∈ I (16)
a subset of binary variables with non-zero coefficients
tsi ≤ di − pim xim ∀i ∈ I (17)
in the objective function, they can in principle be gen- m∈M
eralized for MILP models that have both binary and
xim = 1 ∀i ∈ I (18)
continuous variables in the objective function. The m∈M
xim pim ≤ maxdi − minri ∀m ∈ M (19) model, is highly dependent on the CP package used
i i
i∈I to model the problem because of the differences in
yii + yi i ≥ xim + xi m − 1 constructs available in various modeling languages.
In this paper we use ILOG’s OPL modeling lan-
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
assigned on the same machine should be less than s.t. istart ≥ ri ∀i ∈ I (28)
the difference of the latest due date and the earli- istart ≤ di − pzi ∀i ∈ I (29)
est release date. Constraint (20) is a logical relation-
iduration = pzi ∀i ∈ I (30)
ship between assignment and sequencing variables.
The underlying logic behind this constraint is that i requires T ∀i ∈ I (31)
if order i and i are assigned to machine m then activityHasSelectedResourcei T tm
they must be processed one after the other. Con-
⇔ zi = m ∀i ∈ I m ∈ M (32)
straint (21) is the sequencing constraint in which
U = i∈I maxm∈M pim . It ensures that if the sequenc- zi ∈ M ∀i ∈ I (33)
ing variable yii is one, then order i is processed after istart ∈ ∀i ∈ I (34)
order i. Constraints (22) and (23) are logical cuts. The
iduration ∈ ∀i ∈ I (35)
former is based on the logic that either i is processed
before i or vice versa. The latter ensures that sequenc- Here zi is a variable subscript and represents the
ing variables are zero if i and i are assigned to dif-
machine selected to process order i. Variable sub-
ferent machines. Adding both of these constraints
scripts are powerful modeling constructs and avail-
reduces the computational time needed to solve the
able in most CP software packages. The variable
MILP very significantly. Other logic cuts were also
subscript zi has been used in the objective func-
tried, but they were not as effective as (22) and (23).
tion (27) to calculate the cost of order i (Cizi ) directly.
5.2. CP Model Constraints (28) and (29) ensure that processing of
The same scheduling problem can also be modeled order i begins after release date ri and is completed
using CP. The CP model, in contrast to an MILP before the due date di . The start time of the order is
given by “istart”. The duration of an order depends 5.3. Combined MILP-CP OPL Model
on the machine used to process it. This is enforced in Instead of developing pure MILP and CP models, it is
constraint (30), which uses the variable subscript zi to also possible to write a combined model. OPL allows
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
calculate processing time of order i. Constraint (31) modeling of a problem using a combined MILP-CP
uses a special OPL construct “requires”. It enforces model in which constraints are expressed in both
that order i needs a unary resource from the set of forms. Even though the size of the integrated model
unary resources T . Note that this constraint enforces is larger, it may still perform better in some cases
the assignment of an order to a specific machine because fewer nodes may have to be explored. The
as well as the sequencing of orders that have been solution algorithm for the integrated model primarily
assigned to the same machine. Constraint (32) uses uses the CP solver. However, at each CP node an extra
another OPL-specific boolean function “activityHas- LP relaxation, consisting of all the linear constraints
SelectedResource()” that returns a value of true or in the combined model, is solved to obtain bounds
false. In the context of our scheduling problem this for the objective function (Van Hentenryck 1999). The
function returns a value of “true” if order i is pro- combined MILP-CP OPL model for the scheduling
cessed using the unary resource corresponding to problem at hand can be written as:
the machine m (tm ∈ T ), and false otherwise. Con-
min Cim xim (36)
straint (32) ensures that if order i is processed using i∈I m∈M
the unary resource tm , then the subscript variable zi is s.t.
equal to m. It links the decisions made by the infer-
tsi ≥ ri ∀i∈ I
ence engine behind the construct “requires” and other
tsi ≤ di − pim xim ∀i ∈ I
constraints in the model.
m∈M
It can be clearly seen that variable and constraint
xim = 1 ∀i ∈ I MILP (37)
definitions in the CP model for the scheduling prob-
m∈M
lem at hand are very different from the MILP model.
xim pim ≤ maxdi − minri
This is primarily because specialized CP constructs i i
i∈I
were used to model the problem at hand. However, ∀m ∈ M
there exists a close resemblance in the two model-
istart ≥ ri ∀i ∈ I
ing frameworks. The subscript variable zi in the CP
istart ≤ di − pzi ∀i ∈ I
model is equivalent to the binary variable xim in the
iduration = pzi ∀i ∈ I
MILP model. The start time of order i has the same CP (38)
i requires T ∀i ∈ I
definition in the two models. The sequencing variable activityHasSelectedResourcei T tm
yii in the MILP model has no equivalent variable in
⇔ zi = m ∀i ∈ I m ∈ M
CP model. However, the value of this variable can be
xizi = 1 ∀i ∈ I
obtained from the solution of the CP model. Equiv-
istart= tsi ∀i ∈ I
alence can also be drawn between the constraints in (39)
Ci = Cim xim ∀i ∈ I Linking
the two frameworks. Constraints (28) and (29) are
Constraints
m∈M
equivalent to (16) and (17), respectively. The set of i = Ciz ∀i ∈ I
C i
constraints (30) to (32) play the same role as the tsi ≥ 0 (40)
set of constraints (18), (20), and (21). One major dif-
ference between the two models is that the MILP xim ∈ 0 1 ∀i ∈ I m ∈ M (41)
model is a continuous-time scheduling model. The CP zi ∈ M ∀i ∈ I (42)
model, on the other hand, is a discrete-time schedul-
istart ∈ ∀i ∈ I (43)
ing model (start time is an integer variable in the OPL
framework). iduration ∈ ∀i ∈ I (44)
1 1 15 33 3 3 000 26
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
1 2 15 33 15 19 002 18
2 1 42 77 12 12 004 60
2 2 42 77 172 185 014 44
3 1 72 132 4743 4748 384 101
3 2 72 132 583 606 038 83
4 1 120 165 469918 469938 55354 115
4 2 120 165 5435 5467 928 102
5 1 160 220 55529196 55529254 6885349b 165c
5 2 160 220 1307486 1307532 267387 140
a
Using ILOG Scheduler 4.4 and ILOG Solver 4.4 single processor versions on a dual processor SUN Ultra 60
workstation
b
Computational time for finding the suboptimal solution
c
Suboptimal Solution
data set of the biggest example (Problem 5) could not the possible ways is to use the combined MILP-
be solved to optimality using any of the two meth- CP OPL model that was presented in the previous
ods in reasonable computational time. It should be section. The computational results for solving the set
noted that computational times for MILP model may of ten example problems using the MILP-CP OPL
be possibly reduced further by using more sophis- model are presented in Table 4. Clearly, the MILP-CP
ticated schemes like branch and cut. Similarly, com- OPL model for all the problems is larger than the cor-
putational times for the CP model may be possibly responding CP model. However, it is still smaller than
reduced by developing special branching schemes for the corresponding MILP model because the sequenc-
this example. These strategies are outside the scope of ing variables and constraints were not included in the
this paper. combined MILP-CP OPL model. It can be clearly seen
The advantage of the MILP model is that the val- that all the problems could be solved faster using the
ues of the assignment variables obtained from the combined model. Furthermore, the first data set of the
LP relaxations direct the branch-and-bound search to fifth example problem that was intractable for both
obtain the least-cost assignments. However, schedul- the MILP and CP methods alone could now be solved
ing the orders on each machine is a more daunting to optimality using the combined method.
task. This is because the sequencing constraint (21)
has a big-M form and results in a poor relaxation.
6. MILP/CP Hybrid Model
Furthermore, the sequencing variables do not directly
The scheduling problem at hand can be posed in the
contribute to the objective-function value and the
proposed hybrid MILP/CP modeling framework as
LP relaxation may or may not direct these variables follows,
towards feasibility. On the other hand, the CP meth-
ods use effective constraint-propagation algorithms min Cim xim (45)
i∈I m∈M
for scheduling that are based on the ideas of edge
finding and task intervals (Caseau and Laburthe 1994, s.t. tsi ≥ ri ∀i ∈ I (46)
1996). However, an assignment is chosen using a pre-
tsi ≤ di − pim xim ∀i ∈ I (47)
defined enumeration strategy by excluding the infea- m∈M
sible schedules.
xim = 1 ∀i ∈ I (48)
Clearly, there are complementary strengths of the m∈M
MILP and the CP methods that may possibly be
xim pim ≤ maxdi − minri ∀m ∈ M (49)
combined to tackle more difficult problems. One of i∈I
i i
1 1 38 49 3 3 004 26
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
1 2 38 49 15 19 005 18
2 1 94 120 13 13 010 60
2 2 94 120 175 188 027 44
3 1 159 205 1939 1944 421 101
3 2 159 205 693 716 112 83
4 1 230 286 43882 43902 9159 115
4 2 230 286 3057 3090 558 102
5 1 305 381 4929656 4929686 1373606 158
5 2 305 381 80290 80337 17095 140
a
Using CPLEX 6.5, ILOG Scheduler 4.4, and ILOG Solver 4.4 single processor versions on a dual processor SUN
Ultra 60 workstation
if xim = 1 then zi = m ∀i ∈ I m ∈ M (50) the unary resource corresponding to the machine to
which it has been assigned. These constraints are
istart ≥ ri ∀i ∈ I (51)
equivalent to constraint set (11) of model (M3).
istart ≤ di − pzi ∀i ∈ I (52) Any one of the two proposed algorithms can be
iduration = pzi ∀i ∈ I (53) used to solve this hybrid MILP/CP model for the
scheduling problem at hand. In this section, we use
i requires tzi ∀i ∈ I (54) the decomposition algorithm that was presented in
tsi ≥ 0 (55) Figure 1 for solving the hybrid MILP/CP model.
The reason for choosing this method is that it is
xim ∈ 0 1 ∀i ∈ I m ∈ M (56)
easy to implement and is sufficient to test the poten-
zi ∈ M ∀i ∈ I (57) tial usefulness of the proposed methodology. Details
of the decomposition algorithm in the context of
istart ∈ ∀i ∈ I (58)
the proposed hybrid MILP/CP scheduling model are
iduration ∈ ∀i ∈ I (59) presented in Figure 2. In this algorithm, the relaxed
MILP problem assigns a machine to every order. The
The objective function (45) for the hybrid model
CP feasibility problem then attempts to find a feasi-
is the same as the objective function (15). Con-
ble schedule for this assignment. If a feasible schedule
straints (46) through (49) are the MILP constraints (16)
through (19). These constraints are equivalent to con- can be obtained, then it is the optimal solution. Oth-
straint set (9) of model (M3). Constraint (50) estab- erwise, causes of infeasibilities are inferred as integer
lishes the equivalence between the MILP variable xim cuts to exclude assignments with similar characteris-
and the CP variable zi , and it is represented by (10) in tics. The relaxed MILP problem is resolved to obtain
model (M3). It should be noted that there is no need another assignment. If no other assignment is possi-
to write the equivalence relation in closed form as ble, then the scheduling problem is infeasible.
equations or inequalities. This is in sharp contrast to Integer cuts used in the decomposition algorithm
the set of equivalence relations (39) used in the MILP- are very critical to its efficiency. Instead of using the
CP OPL model. Finally, constraints (51) through (54) integer cuts in (14), the cuts used for the scheduling
are the reduced set of CP constraints derived from the problem at hand are based on the idea that if a set of
original CP constraints (28) through (32) by assuming orders cannot be scheduled on a particular machine,
that the assignment of orders to machines has already then it will not be possible to find a feasible sched-
been made. Note that the CP constraint (32) is no ule for any assignment in which all those orders are
longer needed, and in constraint (54) order i requires assigned to that machine. Note that a cut is generated
min Cim xim
i∈I m∈M No Solution ✲ Infeasible
s.t. (46) to (49), (55) to (56)
xim ≤ Bkm − 1 ✛
k
i∈Im
Imk =
i
xim
k
= 1 Bkm =
Imk
∀m ∈ M k =
1 2 K − 1
Feasible
❄
Optimal Solution Found
for each machine that could not be scheduled suc- In this table, the size of the CP subproblems has
cessfully. The integer cuts for iteration k, which are not been reported as it varies from one iteration
stronger than equation (14), have the following gen- to another, and only the initial size of the relaxed
eral form, MILP problems has been reported. However, the final
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
MILP Size
Major MILP CP
Problem Set Constraints Variables Iterations Cuts Timea Timeb Objective
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
1 1 11 9 2 1 002 000 26
1 2 11 9 1 0 001 000 18
2 1 24 28 13 16 047 005 60
2 2 24 28 1 0 001 001 44
3 1 39 48 31 43 401 017 101
3 2 39 48 1 0 002 000 83
4 1 50 90 18 26 201 024 115
4 2 50 90 1 0 002 002 102
5 1 65 120 31 60 1369 044 158
5 2 65 120 6 6 029 012 140
a
Using CPLEX 6.5 single processor version on a dual processor SUN Ultra 60 workstation
b
Using ILOG Scheduler 4.4 and ILOG Solver 4.4 single processor versions on a dual processor SUN Ultra 60
workstation
MILP and CP to solve problems that are otherwise fewer constraints as compared to the MILP model.
intractable if solved using either of the two meth- However, the computational effort required to solve
ods alone. The class of problems considered in this the problems using any of the two models increases
paper has the characteristic that only a subset of the rapidly with problem size. The MILP-CP OPL model
binary variables have a non zero objective function was larger in size than the CP model but was smaller
coefficients if modeled as an MILP. This class of prob- compared to the MILP model. The combined OPL
lems can be formulated as a hybrid MILP/CP model model could solve most of the problems faster rel-
that involves some of the MILP constraints, a reduced
ative to the MILP and CP models. The scheduling
set of CP constraints, and the equivalence relations
problem was then modeled as a hybrid MILP/CP
between the MILP and the CP variables. To solve
model and solved using the proposed decomposition
this hybrid model, an MILP/CP based decomposition
method and an LP/CP based branch-and-bound algo- algorithm. Computational results indicate that for the
rithm were proposed. Both of these algorithms rely larger instances of the example scheduling problem,
on the same relaxed MILP and feasibility CP prob- the computational times were significantly smaller
lems. The aim of these methods is to combine the compared to MILP, CP, or combined MILP-CP OPL
strength of MILP for proving optimality by using the methods. The example problem demonstrates that the
LP relaxations, and the power of CP for finding fea- proposed methods can be computationally efficient
sible solutions for hard discrete optimization prob- for solving certain problems. It is acknowledged that
lems by using the specialized propagation algorithms. even if a problem falls into the class of problems
It should be noted that the algorithms presented are considered in this paper, the computational efficiency
fairly general and, if needed, alternative methods for of the proposed algorithms depends strongly on the
solving feasibility problems may also be used. nature of the relaxed MILP problem, the feasibility CP
As an application example, a scheduling problem problem, and the cuts that are used in the algorithm.
was considered. The aim of this problem was to find a
least-cost schedule to process a set of orders using dis-
Acknowledgments
similar parallel machines subject to release and due-
The authors would like to thank Greger Ottosson and Erlendur
date constraints. It was shown that this problem can S. Thorsteinsson for interesting discussions about Constraint Pro-
be modeled as an MILP, CP, or a combined MILP- gramming and Iiro Harjunkoski for his help in revising the paper.
CP OPL model (Van Hentenryck 1999). The compu- Financial support from the National Science Foundation under
tational results indicate that the CP model requires Grant CTS-9810182 is gratefully acknowledged.
Balas, E., S. Ceria, G. Cornuejols. 1996. Mixed 0-1 programming by Discrete Applied Mathematics 96–97 395–442.
Lift-and-Project in a Branch-and-Cut framework. Management Hooker, J. N., G. Ottosson, E. S. Thorsteinsson, H. J. Kim. 1999. On
Science 42 1229–1246. integrating constraint propagation and linear programming for
Barnhart, C., E. L. Johnson, G. L. Nemhauser, M. W. P. Savelsberg, combinatorial optimization. Proceedings of the Sixteenth National
P. H. Vance. 1998. Branch and price: column generation for Conference on Artificial Intelligence (AAAI-99), AAAI, The AAAI
huge integer programs. Operations Research 46 316–329. Press/MIT Press, Cambridge, MA. 136–141.
Bockmayr, A., T. Kasper. 1998. Branch-and-infer: a unifying frame- ILOG, Inc. 1999a. CPLEX 6.5 User’s Manual. ILOG, Inc., Incline
work for integer and finite domain constraint programming. Village, NV.
INFORMS J. Computing 10 287–300. ILOG, Inc. 1999b. OPL Studio 2.1 User’s Manual. ILOG, Inc., Incline
Carlier, J., E. Pinson. 1989. An algorithm for solving the job-shop Village, NV.
problem. Management Science 35 164–176. ILOG, Inc. 1999c. Scheduler 4.4 User’s Manual. ILOG, Inc., Incline
Caseau, Y., F. Laburthe. 1994. Improving CLP scheduling with task Village, NV.
intervals. P. Van Hentenryck, ed. Logic Programming: Proceedings ILOG, Inc. 1999d. Solver 4.4 User’s Manual. ILOG Inc., Incline
of the 11th International Conference, MIT Press, Cambridge, MA. Village, NV.
369–383. Le Pape, C. 1994. Implementation of resource constraints in
Caseau, Y., F. Laburthe. 1996. Cumulative scheduling with task ILOG schedule: a library for the development of constraint-
based scheduling systems. Intelligent Systems Engineering 3
intervals. M. Maher, ed. Logic Programming: Proceedings of the
55–66.
1996 Joint International Conference and Symposium, MIT Press,
Cambridge, MA. 363–377. Marriott, K., P. J. Stuckey. 1998. Programming with Constraints. MIT
Press, Cambridge, MA.
Chandru, V., J. Hooker. 1999. Optimization Methods for Logical Infer-
Nemhauser, G. L., L. A. Wolsey. 1988. Integer and Combinatorial Opti-
ence. John Wiley & Sons, Inc., New York.
mization. John Wiley and Sons, Inc., New York.
Colmerauer, A. 1985. Prolog in 10 figures. Communications of the
Ottosson, G., E. S. Thorsteinsson, J. N. Hooker. 2001. Mixed global
ACM 28 1296–1310.
constraints and inference in hybrid CLP–IP solvers. Annals of
Darby-Dowman, K., J. Little. 1998. Properties of some combinato-
Mathematics and Artificial Intelligence. Forthcoming.
rial optimization problems and their effect on the performance
Padberg, M. W., G. Rinaldi. 1991. A branch-and-cut algorithm for
of integer programming and constraint logic programming.
a symmetric travelling salesman polytope. SIAM Review 33
INFORMS J. Computing 10 276–286.
60–100.
Darby-Dowman, K., J. Little, G. Mitra, M. Zaffalon. 1997. Constraint
Proll, L., B. Smith. 1998. Integer linear programming and constraint
logic programming and integer programming approaches and
logic programming approaches to a template design problem.
their collaboration in solving an assignment scheduling prob-
INFORMS J. Computing 10 265–275.
lem. Constraints, An International Journal 1 245–264.
Quesada, I., I. E. Grossmann. 1992. An LP/NLP based branch-and-
Dincbas, M., P. Van Hentenryck, H. Simonis, A. Aggoun, T. Graf, bound algorithm for convex MINLP optimization problems.
F. Berthier. 1988. The constraint logic programming language Computers Chem. Engng. 19 937–947.
CHIP. FGCS-88: Proceedings of International Conference on Fifth
Raman, R., I. E. Grossmann. 1993. Symbolic integration of logic
Generation Computer Systems, Tokyo, Japan. 693–702.
in MILP branch and bound techniques for the synthe-
Geoffrion, A. M. 1972. Generalized Benders decomposition. Journal sis of process networks. Annals of Operations Research 42
of Optimization Theory and Applications 10 237–260. 169–191.
Harjunkoski, I., V. Jain, I. E. Grossmann. 2000. Hybrid mixed- Raman, R., I. E. Grossmann. 1994. Modelling and computational
integer/constraint logic programming strategies for solving techniques for logic based integer programming. Computers
scheduling and combinatorial optimization problems. Comp. Chem. Engng. 18 563–578.
Chem. Engng. 24 337–343. Rodosek, R., M. G. Wallace, M. T. Hajian. 1999. A new approach
Heipcke, S. 1999a. An example of integrating constraint program- to integrating mixed integer programming and constraint logic
ming and mathematical programming. Electronic Notes in Dis- programming. Annals of Operational Research 86 63–87.
crete Mathematics 1. Smith, B. M., S. C. Brailsford, P. M. Hubbard, H. P. Williams. 1997.
Heipcke, S. 1999b. Comparing constraint programming and mathe- The progressive party problem: integer linear programming
matical programming approaches to discrete optimisation-the and constraint programming compared. Constraints, An Inter-
change problem. J. of Operational Research Society 50 581–595. national Journal 1 119–138.
Tsang, E. P. K. 1993. Foundations of Constraint Satisfaction. Academic Van Hentenryck, P. 1999. The OPL Optimization Programming Lan-
Press, Inc., San Diego, CA. guage. MIT Press, Cambridge, MA.
Van Hentenryck, P. 1989. Constraint Satisfaction in Logic Program- Wallace, M., S. Novello, J. Schimpf. 1997. ECLiPSe: a platform for
ming. MIT Press, Cambridge, MA. constraint logic programming. ICL Systems Journal 12 159–200.
Downloaded from informs.org by [128.122.253.212] on 07 June 2015, at 20:49 . For personal use only, all rights reserved.
Accepted by John N. Hooker; received October 1999; revised July 2000, Februray 2001; accepted March 2001.