Advanced Linear Programming 1

ADVANCED LINEAR PROGRAMMING
4
• PERSONNEL
SCHEDULING • BINARY
• RADIATION • GOAL
PART I PART II
THERAPY • NON-LINEAR
• OIL REFINING PROGRAMMING
• INTEGER 1
1
Dr. Faheem Qaisar Jamal, School of Industrial Engineering (SIE), CAE, NUST 1
More Linear Programming Models
Dr. Faheem Qaisar Jamal, School of Industrial Engineering (SIE), CAE, NUST
MIT and James Orlin © 2003 2
 Overview
– 5 in 7 scheduling problem
• The model
Personnel Scheduling • Practical enhancements or
Programming Models modifications
• Two non-linear objectives that can be
made linear
• A non-linear constraint that can be
made linear
Overview
 5 in 7 scheduling problem
– The model
Scheduling Postal Workers
 Each postal worker works for 5 consecutive days,
followed by 2 days off, repeated weekly.
Day Mon Tues Wed Thurs Fri Sat Sun
Demand 17 13 15 19 14 16 11
 Minimize the number of postal workers (for the

time being, we will permit fractional workers on
each day.)
Formulating as an LP
 Select the decision variables
– Let x1 be the number of workers who
start working on Monday, and work till
Friday
– Let x2 be the number of workers who
start on Tuesday …
– Let x3, x4, …, x7 be defined similarly.
Day Mon Tues Wed Thurs Fri Sat Sun
The linear
Demand 17 program
13 15 19 14 16 11
Minimize z = x1 + x2 + x3 + x4 + x5 + x6 + x7
subject to x1 + x4 + x5 + x6 + x7  17
x1 + x2 + x5 + x6 + x7  13
x1 + x2 + x3 + x6 + x7  15
x1 + x2 + x3 + x4 + x7  19
x1 + x2 + x3 + x4 + x5  14
x2 + x3 + x4 + x5 + x6  16
x3 + x4 + x5 + x6 + x7  11
xj  0 for j = 1 to 7
On the selection of decision variables
 Would it be possible to have yj be the
number of workers on day j?
– Workers on day j is at least dj.
– Each worker works 5 days on followed by 2
days off.
 Conclusion: sometimes the decision
variables incorporate constraints of the
problem.
– Hard to do this well, but worth keeping in mind
– We will see more of this in integer
programming.
Some Enhancements of the Model
 Suppose that there was a pay differential. The
cost of workers who start work on day j is cj
per worker.
Minimize z = c1 x1 + c2 x2 + c3 x3 + … + c7 x7
Overview
– The model
– Practical enhancements or modifications
Some Enhancements of the Model
 Suppose that one can hire part time workers

(one day at a time), and that the cost of a part
time worker on day j is PTj.
 Let yj = number of part time workers on day j
What is the Revised Linear Program?
Minimize z = z = c1 x1 + c2 x2 + c3 x3 + … + c7 x7
subject to x1 + x4 + x5 + x6 + x7  17
x1 + x2 + x5 + x6 + x7  13
x1 + x2 + x3 + x6 + x7  15
x1 + x2 + x3 + x4 + x7  19
x1 + x2 + x3 + x4 + x5  14
x2 + x3 + x4 + x5 + x6  16
x3 + x4 + x5 + x6 + x7  11
xj  0 for j = 1 to 7
Minimize z = c1 x1 + c2 x2 + c3 x3 + … + c7 x7
+ PT1 y1 + PT2 y2 + … + PT7 y7
subject to x1 + x4 + x5 + x6 + x7 + y1  17
x1 + x2 + x5 + x6 + x7 + y2  13
x1 + x2 + x3 + x6 + x7 + y3  15
x1 + x2 + x3 + x4 + x7 + y4  19
x1 + x2 + x3 + x4 + x5 + y5  14
x2 + x3 + x4 + x5 + x6 + y6  16
x3 + x4 + x5 + x6 + x7 + y7  11
xj  0 , yj  0 for j = 1 to 7
Another Enhancement
 Suppose that the number of workers required on
day j is dj. Let yj be the number of workers on
day j.
 What is the minimum cost schedule, where the

“cost” of having too many workers on day j is -
fj(yj – dj), which is a non-linear function?
 NOTE: this will lead to a non-linear program, not

a linear program.
 We will let sj = yj – dj be the excess number of

workers on day j.
What is the Revised Linear Program?
Minimize z = x1 + x2 + x3 + x4 + x5 + x6 + x7
subject to x1 + x4 + x5 + x6 + x7  17
x1 + x2 + x5 + x6 + x7  13
x1 + x2 + x3 + x6 + x7  15
x1 + x2 + x3 + x4 + x7  19
x1 + x2 + x3 + x4 + x5  14
x2 + x3 + x4 + x5 + x6  16
x3 + x4 + x5 + x6 + x7  11
xj  0 for j = 1 to 7
Minimize
z = f1(s1) + f2(s2) + f3(s3) + f4(s4) + f5(s5) + f6(s6) + f7(s7)
subject to x1 + x4 + x5 + x6 + x7 - s1 = 17
x1 + x2 + x5 + x6 + x7 - s2 = 13
x1 + x2 + x3 + x6 + x7 - s3 = 15
x1 + x2 + x3 + x4 + x7 - s4 = 19
x1 + x2 + x3 + x4 + x5 - s5 = 14
x2 + x3 + x4 + x5 + x6 - s6 = 16
x3 + x4 + x5 + x6 + x7 - s7 = 11
xj  0 , sj  0 for j = 1 to 7
Overview
– The model
– Two non-linear objectives that can be made
linear
A non-linear objective that often can be made
linear.
Suppose that one wants to minimize the

maximum of the slacks, that is
minimize z = max (s1, s2, …, s7).
This is a non-linear objective.

But we can transform it, so the problem becomes
an LP.
Minimize z
z  sj for j = 1 to 7.
subject to x1 + x4 + x5 + x6 + x7 - s1 = 17
x 1 + x2 + x5 + x6 + x7 - s2 = 13
x1 + x2 + x3 + x6 + x7 - s3 = 15
x1 + x2 + x3 + x4 + x7 - s4 = 19
x1 + x2 + x3 + x4 + x5 - s5 = 14
x2 + x3 + x4 + x5 + x6 - s6 = 16
x3 + x4 + x5 + x6 + x7 - s7 = 11
xj  0 , sj  0 for j = 1 to 7
The new constraint ensures that z  max (s1, …, s7)
The objective ensures that z = sj for some j.
Another non-linear objective that often can be
made linear.
Suppose that the “goal” is to have dj workers on
day j. Let yj be the number of workers on day j.
Suppose that the objective is

minimize Si | yj – dj |
This is a non-linear objective.

But we can transform it, so the problem becomes an LP.
Minimize Sj zj
z j  d j - yj for j = 1 to 7.
zj  yj - dj for j = 1 to 7.
subject to x1 + x4 + x5 + x6 + x7 = y1
x 1 + x2 + x5 + x6 + x7 = y2
x1 + x2 + x3 + x6 + x7 = y3
x1 + x2 + x3 + x4 + x7 = y4
x1 + x2 + x3 + x4 + x5 - = y5
x2 + x3 + x4 + x5 + x6 = y6
x3 + x4 + x5 + x6 + x7 = y7
xj  0 , yj  0 for j = 1 to 7
The new constraints ensure that zj  | yj – dj | for each j.
The objective ensures that zj = | yj – dj | for each j.
Overview
– The model
– Two non-linear objectives that can be made
linear
– A non-linear constraint that can be made linear
A ratio constraint:
Suppose that we need to ensure that at least 30% of the
workers have Sunday off.
How do we model this?
(x1 + x2 )/x1 + x2 + x3 + x4 + x5 + x6 + x7  .3
(x1 + x2 )  .3 x1 + .3 x2 + .3 x3 + .3 x4 + .3 x5 + .3 x6 + .3 x7
-.7 x1 - .7 x2 + .3 x3 + .3 x4 + .3 x5 + .3 x6 + .3 x7 <= 0
Other enhancements
 Require that each shift has an integral
number of workers
– integer program
 Consider longer term scheduling
– model 6 weeks at a time
 Consider shorter term scheduling
– model lunch breaks
 Model individual workers
– permit worker preferences
Radiation Treatments – Overview of how math
Problem programming arises in radiation
treatment design
– Formulation as math programs
– Discussion issues
Radiation Therapy Overview
 High doses of radiation (energy/unit mass) can kill cells and/or
prevent them from growing and dividing
– True for cancer cells and normal cells
 Radiation is attractive because the repair mechanisms for cancer
cells is less efficient than for normal cells
Radiation Therapy Overview
 Recent advances in radiation therapy now make it possible to
– map the cancerous region in greater detail
– aim a larger number of different beamlets with greater specificity
 This has spawned the new field of tomotherapy
 “Optimizing the Delivery of Radiation Therapy to Cancer patients,”
by Shepard, Ferris, Olivera, and Mackie, SIAM Review, Vol 41, pp
721-744, 1999.
 Also see http://www.tomotherapy.com/
Conventional Radiotherapy
Relative Intensity of Dose Delivered
Relative Intensity of Dose Delivered
 In conventional radiotherapy
– 3 to 7 beams of radiation
– radiation oncologist and physicist work together to determine a
set of beam angles and beam intensities
– determined by manual “trial-and-error” process
Goal: maximize the dose to the tumor while minimizing
dose to the critical area
Critical Area
Tumor area
With a small number of beams, it is difficult to achieve these goals.
Recent Advances
 More accurate map of tumor area
– CT -- Computed Tomography
– MRI -- Magnetic Resonance Imaging
 More accurate delivery of radiation

– IMRT: Intensity Modulated Radiation Therapy
– Tomotherapy
Tomotherapy: a diagram
Radiation Therapy: Problem Statement
 For a given tumor and given critical areas

 For a given set of possible beamlet origins and angles
 Determine the weight of each beamlet such that:
– dosage over the tumor area will be at least a target level gL .
– dosage over the critical area will be at most a target level gU.
Display of radiation levels
Linear Programming Model
 First, discretize the space

– Divide up region into a 2D (or 3D) grid of pixels
More on the LP
 Create the beamlet data for each of

p = 1, ..., n possible beamlets.
 Dp is the matrix of unit doses delivered by beam p.
Dijp = unit dose

delivered to pixel (i,j)
by beamlet p
Linear Program
 Decision variables w = (w1, ..., wp)

 wp = intensity weight assigned to beamlet p
for p = 1 to n;
 Dij = dosage delivered to pixel (i,j)
Dij   p1 D w p
n p
ij
An LP model took 4 minutes to
minimize  (i, j)
Dij
solve.
Dij   p1 D w p
n p
ij
Dij   L for ( i , j )  T
Dij   U for ( i , j )  C
wp  0 for all p
In an example reported in the paper, there were more than 63,000 variables,
and more than 94,000 constraints (excluding upper/lower bounds)
What to do if there is no feasible solution?
 Use penalties: e.g., Dij  gL – yij

and then penalize y in the objective.
 Consider non-linear penalties (e.g., quadratic)
 Consider costs that depend on damage rather than on radiation
 Develop target doses and penalize deviation from the target
Optimal Solution for the LP
An Optimal Solution to an NLP
Further considerations
 Minimize damage to critical tissue
 Maximize damage to tumor cells
 Minimize time to carry out the dosage
 LP depends on the technology
Oil Refining Programming Model
Problem Statement
A manager of an oil refinery has 8 million barrels of crude oil ‘A’
and 5 million barrels of crude oil ‘B’ allocated for production
during the coming month. These resources can be used to
make either gasoline, which sells for $38 per barrel or home
heating oil, which sells for $33 per barrel. There are three
production processes with following characteristics:-
Process 1 Process 2 Process 3
Input Crude A 3 1 5
Input Crude B 5 1 3
Output Gasoline 4 1 3
Output Heating Oil 3 1 4
Cost ($) 51 11 40
Maximize the revenue of the company
Solution of Oil Refinery problem
DVs X1= Unit quantity with process 1
A manager of an oil refinery has 8 million barrels of crude oil
‘A’ and 5 million barrels of crude oil ‘B’ allocated for X2=Unit quantity with process 2
production during the coming month. These resources can X3 =Unit quantity with process 3
be used to make either gasoline, which sells for $38 per
barrel or home heating oil, which sells for $33 per barrel. Selling: 38 x (4X1+X2+3X3 ) + 33 x (3X1+X2+4X3 )
There are three production processes with following
= 251X1+71X2+246X3 )
characteristics:-
Cost : 51X1+11X2+40X3
Process 1 Process 2 Process 3
Input Crude A 3 1 5 Net Revenue: Sale – Cost
Input Crude B 5 1 3 = (251X1+71X2+246X3) – (51X1+11X2+40X3)
Output Gasoline 4 1 3 = 200X1+60X2+206X3
Output Heating Oil 3 1 4
Max. Objective, z: 200X1 + 60X2 + 206 X3
Cost ($) 51 11 40
Constraints:
Crude Oil A: 3X1+X2+5X3  8 x 106
Maximize the revenue of the company
Crude Oil B: 5X1+X2+3X3  5 x 106
X1 ≥ 0, X2 ≥ 0, X3 ≥ 0
(Binary)
We use Branch and Bound Method to solve ILP
Questions
1

Advanced Linear Programming 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Advanced Linear Programming 1

Uploaded by

Copyright:

Available Formats

ADVANCED LINEAR PROGRAMMING

Day Mon Tues Wed Thurs Fri Sat Sun

 Minimize the number of postal workers (for the

 Suppose that one can hire part time workers

 Let yj = number of part time workers on day j

 What is the minimum cost schedule, where the

 NOTE: this will lead to a non-linear program, not

 We will let sj = yj – dj be the excess number of

Suppose that one wants to minimize the

This is a non-linear objective.

Suppose that the objective is

This is a non-linear objective.

Relative Intensity of Dose Delivered

Relative Intensity of Dose Delivered

With a small number of beams, it is difficult to achieve these goals.

 More accurate delivery of radiation

 For a given tumor and given critical areas

 First, discretize the space

 Create the beamlet data for each of

Dijp = unit dose

 Decision variables w = (w1, ..., wp)

 Dij = dosage delivered to pixel (i,j)

 Use penalties: e.g., Dij  gL – yij

Maximize the revenue of the company

You might also like