ILOG OPL Studio

Technical overview
Sofiane Oussedik
[email protected]
1
Outline
ILOG OPL Studio

• Overview
• Models and Data
• Linear Programming Example
• Constraint Programming Example
• Scheduling Example
• Data Management Features
• Iterative Applications
• OPL Component Libraries
2
Optimization Development
ILOG OPL Studio

Business
Process
Expert Programming
Expert

ILOG OPL Studio reduces time to market

for developing optimization applications

3
ILOG Optimization Suite

ILOG
OPL ILOG ILOG ILOG
Studio Scheduler Dispatcher Configurator

ILOG
C JConfigurator
API
ILOG
VB ILOG Solver & ILOG JSolver
CPLEX
API
ILOG Concert Technology (C++ & Java)

A Common API for CPLEX and Solver

4
Inside OPL Studio
ILOG OPL Studio

• OPL Optimization Programming Language

• A language for representing optimization problems
• Has advanced types to allow better organization of data
• Supports constraint programming, linear and integer
programming, and scheduling problems
• Database and Microsoft Excel connectivity
• OPLScript for iterative solving and hybrid optimization

• Graphical User Interface for Optimization Problems

• Text editor with keyword colors for entering problems and data
• Visualizations of data and solutions
• Menus/buttons for controlling optimization
• Online help for OPL language
5
OPL Component Libraries
ILOG OPL Studio

• API's for embedding OPL models and OPL scripts

• C++
• Microsoft COM / .NET
• Visual Basic
• Visual Basic for Applications (Excel, Access, etc.)
• Java
• Web
• ASP, JSP

• Links with ILOG CPLEX, Solver, and Scheduler

6
 OPL features

for development
7
What Are Models?
Linear programming

• A data-independent abstraction of a problem

• OPL lets you write down the mathematical
representation of a model separately from the data

Project
Model Data

One
Problem
Instance
8
A Production Planning Example
Linear programming

• A manufacturer wants to sell a product

• The product can be made either
• Inside the factory
• Scarce resources are required to build the product
• There is a cost per unit to manufacture the product
• Outside the factory
• There is a cost per unit to purchase the product

• All demand must be satisfied

• The goal is to minimize cost
9
Data Declarations
Linear programming
• Sets of products and resources
enum Products ...;
enum Resources ...;
• Number of units of each resource needed to produce one unit
of each product
float+ consumption[Products, Resources] = ...;
• Total number of available resources
float+ capacity[Resources] = ...;
• Number of units in demand for each product
float+ demand[Products] = ...;
• Cost per unit of inside and outside production
float+ insideCost[Products] = ...;
float+ outsideCost[Products] = ...;
10
Products Could Be Jewelry
Linear programming
• Products and Resources
Products = { rings earrings };
Resources = { gold diamonds };
• Consumption
• A ring requires 3 units of gold and 1 diamond
• A set of earrings requires 2 units of gold and 2 diamonds
consumption = [ [3, 1], [2, 2] ];
• Capacity (Available units of gold and diamonds)
capacity = [ 130, 180 ];
• Demand (Number of rings and earrings)
demand = [ 100, 150 ];
• Costs (per unit for rings and earrings)
insideCost = [ 250, 200 ];
outsideCost = [ 260, 270 ];
11
Products Could Be Pasta
Linear programming
• Products and Resources
Products = { kluski capellini fettucine };
Resources = { flour eggs };
• Consumption
• Kluski requires 0.5 units of flour and 0.2 eggs
• Capellini requires 0.4 units of flour and 0.4 eggs
• Fettucine requires 0.3 units of flour and 0.6 eggs
consumption = [ [0.5, 0.2], [0.4, 0.4], [0.3, 0.6] ];
• Capacity (Available units of flour and eggs)
capacity = [20, 40];
• Demand (Number of each pasta needed)
demand = [100, 200, 300];
• Costs (per unit for each pasta)
insideCost = [0.6, 0.8, 0.3];
outsideCost = [0.8, 0.9, 0.4]; 12
Problem Model Is Identical
Linear programming
enum Products ...;
enum Resources ...;

float+ consumption[Products, Resources] = ...;

float+ capacity[Resources] = ...;
float+ demand[Products] = ...;
float+ insideCost[Products] = ...; Data
float+ outsideCost[Products] = ...;

var float+ inside[Products];

var float+ outside[Products]; Decision
Variables

13
Problem Model Is Identical (2)
Linear programming
minimize
sum(p in Products)
(insideCost[p] * inside[p] +
Objective
outsideCost[p]* outside[p] ) Function

subject to {
forall(r in Resources)
sum(p in Products)
consumption[p, r]*inside[p] <= capacity[r];

forall(p in Products)
inside[p] + outside[p] >= demand[p];
};

Constraints
14
Demo of production.prj
Linear programming

• Run production.prj
• Display inside and outside production

15
Linear and Integer Programming
Linear programming
• OPL can represent linear and mixed integer programming
problems
• Objective functions and constraints are linear
• E.g., 5 x + 3 y - 4 z
• Variables can be floating point or integer valued
• Constraints are inequalities or equalities ( ≤, ≥, =)
• OPL Studio can solve
• Linear Programs with any of the CPLEX algorithms
• Mixed Integer Programs using the CPLEX MIP algorithm
• Mixed Integer Programs by using constraint programming-based
search
• OPL provides access to all CPLEX algorithmic settings

16
Map Coloring Example
Constraint programming

• Given a list of countries

enum Country {Belgium,Denmark,France,Germany,
Netherlands,Luxembourg};

• A set of colors to assign to countries on the map

enum Colors {blue,red,yellow,gray};

• Want to decide how to assign the colors to the

countries so that no two bordering countries have
the same color
var Colors color[Country];

The decision variables

are values from a set
17
Constraint Programming Model
Constraint programming
enum Country {Belgium,Denmark,France,Germany,
Netherlands,Luxembourg};
enum Colors {blue,red,yellow,gray}; Data
var Colors color[Country]; Decision
solve { Find all Solutions Variables

color[France] <> color[Belgium];

color[France] <> color[Luxembourg];
color[France] <> color[Germany];
color[Luxembourg] <> color[Germany];
color[Luxembourg] <> color[Belgium]; Constraints
color[Belgium] <> color[Netherlands];
color[Belgium] <> color[Germany];
color[Germany] <> color[Netherlands];
color[Germany] <> color[Denmark];
};
18
Demo of map.mod
Constraint programming

• Run map.mod
• Show all solutions
• Rerun to show the propagation
• Select Execution > Browse Active Model
• Right-click on Variables > color and select display domain
• Select Debug > Stop at Choice Point
• Step through the model
• Disable Debug > Stop at Choice Point when finished

19
Visualizing Solutions
Constraint programming

• Develop application-specific graphical output using

the drawing board
• Lines
• Polygons
• Arcs, Circles
• Text labels
• Graphics can be added to the search procedure
• Draw an object when variables are fixed to values
• Graphics are updated automatically during the search
procedure

20
Demo of mapgr.prj

• Run mapgr.prj
• Show all solutions

• Graphics help illustrate how many

neighbors constrain each country

21
Constraint Programming Models
Constraint programming
• OPL allows you to represent problems using
constraint programming features
• Variables can be set to a value from a set
• A variable can index into an array ( y[x[k]] )
• Constraint relations can be strict inequalities
(<, >, < >)
• Logical conditions can be modeled
x < 4 => y > 5
• Global constraints can be written
alldifferent(x)
• Constraints can have values (meta constraints)
sum (j in myset) (x[j] > 5) = 3

22
CP Solutions
Constraint programming

• OPL Studio uses ILOG Solver to solve

constraint programming problems
• Default search strategy need not be programmed
• Users can program their own search strategies
• OPL Studio allows debugging of search
strategies
• Users can visualize the values of their variables
• Users can step through a search procedure
• The search tree can be visualized
23
Warehouse Assignment

• Want to assign S stores to W warehouses.

The problem is as follows:
• The cost of assigning store s to warehouse w is
given by the array element supplyCost[s,w].
• Each warehouse w can have at most
capacity[w] stores assigned to it.
• There is a fixed cost fixed=30 for opening up
each warehouse.

24
Warehouse assignment: MIP

var int open[Warehouses] in 0..1;

var int supply[Stores,Warehouses] in 0..1;

minimize
sum(w in Warehouses) fixed * open[w] +
sum(w in Warehouses, s in Stores)
supplyCost[s,w] * supply[s,w]
subject to {
forall(s in Stores)
sum(w in Warehouses) supply[s,w] = 1;
forall(w in Warehouses, s in Stores)
supply[s,w] <= open[w];
forall(w in Warehouses)
sum(s in Stores) supply[s,w] <= capacity[w];
};

25
Warehouse assignment: CP

var int open[Warehouses] in 0..1;

var Warehouses supplier[Stores];
var int cost[Stores] in 0..maxCost;

minimize
sum(s in Stores) cost[s] +
sum(w in Warehouses) fixed * open[w]
subject to {
forall(s in Stores)
cost[s] = supplyCost[s,supplier[s]];
forall(s in Stores )
open[supplier[s]] = 1;
forall(w in Warehouses)
sum(s in Stores) (supplier[s] = w) <= capacity[w];
};

search {
forall(s in Stores ordered by decreasing regretdmin(cost[s]))
tryall(w in Warehouses ordered by increasing supplyCost[s,w])
supplier[s] = w;
}; 26
MIP versus CP formulation

• Decision variables
• The constraint programming formulation has 2S+W decision variables.
• The mixed integer formulation has SW+W decision variables.
• The CP formulation has a decision variable over a finite set of values to
represent the cost of shipping for store s.
• The MIP formulation represents the cost of shipping for store s as an
implied expression.
sum (w in Warehouses) supplyCost[s,w] * supply[s,w]

• Expressions
• The CP formulation uses expressions of the form open[supplier[s]],
which uses a decision variable to index into another decision variable.
• The CP formulation uses the expression (supplier[s] = w) that
evaluates to a 0/1 value.

27
CP includes search!

search {
forall(s in Stores ordered by decreasing regretdmin(cost[s]))
tryall(w in Warehouses ordered by increasing supplyCost[s,w])
supplier[s] = w;
};

• cost[s] can only take on values from supplyCost[s,w] for the set of open
warehouses w
regretdmin = (second lowest value) - (lowest value)

• Pick the store with the largest regret, then pick the warehouse with the smallest
cost
• Then open that warehouse
28
But Which is BETTER ??

• It depends upon the data

• It depends on the search strategy
• It depends on the combinatorial nature of
the problem

• For general applications, you need tools

that allow you to try both methodologies!

29
Scheduling Example (Part 1)
Scheduling
• Have to schedule a set of tasks to build a house
enum Tasks { masonry, carpentry, plumbing,
ceiling, roofing, painting,
windows, facade, garden, moving };
• Tasks require a given amount of time to be completed
int duration[Tasks] = [7,3,8,3,1,2,1,2,1,1];
int totalDuration =
sum(t in Tasks) duration[t];
scheduleHorizon = totalDuration;
• Some tasks require other tasks to be completed
task[masonry] precedes task[ceiling];
task[carpentry] precedes task[roofing];
task[ceiling] precedes task[painting];
...
30
Scheduling Example (Part 2)
Scheduling

• A group of workers must build the house,

but each cannot perform certain tasks
enum Workers { joe, jack, jim };
{Workers} cannotperform[Tasks] = #[
masonry: { jim }, carpentry: { jack },
plumbing: { joe, jim }, ceiling: { jack },
roofing: { jack }, painting: { joe },
windows: { jack }, facade: { jim },
garden: {}, moving: { jack } ]#;

• Goal is to schedule the tasks and the

workers and to minimize the maximum
amount of time any worker is on duty
31
OPL Scheduling Model (Part 1)
Scheduling
enum Tasks { masonry, carpentry, plumbing,
ceiling, roofing, painting,
windows, facade, garden, moving };
int duration[Tasks] = [7,3,8,3,1,2,1,2,1,1];
int totalDuration = sum(t in Tasks) duration[t];
scheduleHorizon = totalDuration; Data
enum Workers { joe, jack, jim };
{Workers} cannotperform[Tasks] = #[
masonry: { jim }, carpentry: { jack },
plumbing: { joe, jim }, ceiling: { jack }, Decision
roofing: { jack }, painting: { joe },
windows: { jack }, facade: { jim },
Variables
garden: {}, moving: { jack } ]#;

Activity task[t in Tasks](duration[t]);

var int durationWorkers[Workers] in 0..totalDuration;
Activity attendance[w in Workers](durationWorkers[w]);
UnaryResource worker[Workers];
AlternativeResources s(worker);
32
OPL Scheduling Model (Part 2)
Scheduling
minimize max(w in Workers) durationWorkers[w]
Objective
subject to {
task[masonry] precedes task[carpentry];
Function
task[masonry] precedes task[plumbing];
task[masonry] precedes task[ceiling]; task[carpentry] precedes task[roofing];
task[ceiling] precedes task[painting]; task[roofing] precedes task[windows];
task[roofing] precedes task[facade]; task[plumbing] precedes task[facade];
task[roofing] precedes task[garden]; task[plumbing] precedes task[garden];
task[windows] precedes task[moving]; task[facade] precedes task[moving];
task[garden] precedes task[moving]; task[painting] precedes task[moving];

forall(t in Tasks) task[t] requires s;

forall(t in Tasks) forall(w in cannotperform[t])

not activityHasSelectedResource(task[t],s,worker[w]);

forall(t in Tasks) forall(w in Workers)

activityHasSelectedResource(task[t],s,worker[w]) =>
attendance[w].start <= task[t].start &
attendance[w].end >= task[t].end;
};

Constraints 33
Demo of house3.mod
Scheduling

• Run house3.mod
• Show the Gantt chart for task activities and
worker resources

34
Scheduling Models
Scheduling
• OPL supports modeling entities for scheduling
problems
• Activities
• Have durations
• Can be breakable
• Constraints to state that activities precede other activities
• Resources
• Unary Resources
• Discrete Resources
• Reservoirs
• State Resources with Transition Times
• Can also add other kinds of constraints using
constraint programming
35
Structured Data
Data Management
• Given a set of cities
enum Cities { MIA, EWR, SFO, BOS };

• Origin/destination pairs can be represented as

struct Pair { Cities o; Cities d; };

• A set of pairs would be represented as

setof(Pair) odpairs =
{<MIA,EWR>,<MIA,SFO>,<SFO,BOS>,<EWR,SFO>};

• The origins can be computed as

setof(Cities) origins = { o | <o,d> in odpairs};

• A decision variable array over the set of odpairs

would be written as:
var float+ flow[odpairs];
36
Database connectivity via SQL
Data Management
• Reading
struct Precedence {
string before; //Task
string after; //Task
};
setof(Precedence) precedences from
DBread(db, "select * from PRECEDENCE");
• Writing
struct Schedule {
string task;
int startTime;
int endTime;
};
setof(Schedule) resultSet = { /* Set of results */ };
DBupdate(db, "insert into Result
(task, startTime, endTime)
values (?, ?, ?)")(resultSet); 37
Excel connectivity
Data Management

• Reading
setof(int) TimePeriods from Time avail
SheetRead(sheetData,"Time"); 1 40
2 40
float+ avail[TimePeriods] from 32
3
SheetRead(sheetData,"avail"); 4 40
setof(string) Products from
SheetRead(sheetData,"Product");
float+ revenue[Products,TimePeriods] from
SheetRead(sheetData,"Revenue");
Revenue 1 2 3 4

• Writing bands
coils
25
30
26
35
27
37
27
39

var float+ Make[Products,TimePeriods];

SheetWrite(sheetResult,"A2:D3")(Make);

• Use named ranges or sheet references 38

OPLScript
Iterative Applications

• Scripting language to control execution of

OPL models and solution methods
• Applications:
• Solve a sequence of related models
• Solve a model with varying data
• Complex decomposition strategies
• Hybrid optimization

39
OPLScript Example
Iterative Applications
Model produce("mulprod.mod","mulprod.dat") editMode;
import enum Resources produce.Resources;
int+ capFlour := produce.capacity[flour];
Declare Model
forall(i in 1..4) {
produce.capacity[flour] := capFlour;
produce.solve(); Set data, then
cout << "Flour capacity" << capFlour << Solve!
"Objective Function: " <<
produce.objectiveValue() << endl;

Basis b(produce); Output Answer

produce.reset();
produce.setBasis(b);
capFlour := capFlour + 1; Try additional
capacity, using
} advanced basis
40
 OPL features

for deployment
41
Deployment Options
OPL Component Libraries
C
O
M OPL Generate Application
P
model compiled Read compiled model with
I model VB, Java, C++, Web Fixed
L Model
E

I
N
T
Application OPL OPL
E Read Read script
R reads interpreted Script model
VB, Java, C++, Web
P scripts and models
R from files
E
T 42
Excel Demo
Steel Production Spreadsheet Application

• Open steelt2D.xls
• Show optimization
• Show how if numbers change, answer
changes

43
Visual Basic Example
OPL Component Libraries
• Initialize solver
Dim solver As COPLsolver
Set solver = New COPLsolver

• Model and data loaded from file or memory

Call solver.loadCompiledModelFileAndDataFile
("mulprod.opl", "mulprod.dat", 1)

• Modeling entities are accessed via strings

Dim capacity As IOPLarray
Set capacity = solver.getArray("capacity")
Dim resources As IOPLenum
Set resources = solver.getEnum("Resources")
Dim flour As IOPLenumValue
Set flour = resources.getValue("flour")
Dim capFlour As IOPLint
Set capFlour = capacity.getInt(flour) 44
C++ and Java almost identical
OPL Component Libraries
• Initialize solver
• C++
OPLsolver solver;

• Java
OPLsolver solver = new OPLsolver();

• Load model and data from file or memory

solver.loadCompiledModelFileAndDataFile("mulprod.opl",
"mulprod.dat", 1);

• Modeling entities are accessed via strings

OPLarray capacity = solver.getArray("capacity");
OPLenum resources = solver.getEnum("Resources");
OPLenumValue flour = resources.getValue("flour");
OPLint capFlour = capacity.getInt(flour);
45
Web JSP Demo
Steel Production Spreadsheet Application

• Open Staffing Demo

• Show optimization

46
OPL Studio Key Features
Summary
• Powerful Modeling Language (OPL) for
• Constraint Programming
• Scheduling
• Linear and Mixed Integer Programming

• Graphical User Interface for

• Entering Models and Data
• Organizing Projects
• Visualizing data and solutions
• Controlling optimization

• Linked with ILOG optimization tools

• ILOG Solver, Scheduler and CPLEX
• Database access, Spreadsheet access and Scripting
• OPL Component Libraries accelerate application deployment
47
Rapid Application Deployment
Summary
• Develop your model in OPL Studio, maintaining
model/data separation
• Refine your algorithm with OPL search strategies
• Use visualization to enhance understanding
• Step through search procedures
• Use iterative or hybrid approaches
• Incorporate model via OPL Component Libraries
• Use C++, Visual Basic, Java, ASP, JSP
• Integrate external data
• Use the answer wherever needed
• Links with ILOG CPLEX, Solver, Scheduler
48
Get more info

Web seminar Series: soon on webseminar.ilog.com

• Overview of CPLEX
• Building CPLEX applications using OPL Studio
• Getting started with OPL Studio

Email: [email protected]

Visit: http://www.ilog.com

49