Optimization of Transportation Problem With Computer Aided Linear Programming
Optimization of Transportation Problem With Computer Aided Linear Programming
ABSTRACT
Theory of Optimizations is to make a better use of resources and existing technology at the best possible
way. One way of optimization is to have a better planning and put it into practice. Precise and quantitative
models, and advanced mathematical tools such as linear programming allows for provable and reproducible
conclusions. Modern computing equipment is suited to put such methods into practice. At present,
mathematical methods contribute, in particular, to the solution of various problems of operational planning
for transportation problems. Most of these transportation problems are often modeled in linear
programming method or in integer programming method. In this paper we investigate these two models for
a real world application of a transportation problem that involves transporting iron ore from two iron ore
mines to three steel plants using linear programming method and integer programming method. The models
are then further applied into the computer to do the analytical task to generate an optimized solution that
minimizes the transportation cost. We have compared the results to find out the most practical model for a
real world situation and how significant the difference would be. Then finally we are able to make a
conclusion on which model would generate a feasible optimal solution for transportation problems and
what are the differences.
KEYWORDS
linear programming, integer programming, optimization, transportation problem
LP by just inverting A. Usually A has more Basic solutions represent extreme boundary
columns than rows, and Ax=b is therefore quite points of the feasible region defined by Ax = b, x
likely to be under-determined, leaving great >= 0, and the simplex method can be viewed as
latitude in the choice of x with which to moving from one such point to another along the
minimize cx[5]. edges of the boundary. Barrier or interior-point
methods, by contrast, visit points within the
The word "Programming" is used here in the interior of the feasible region. These methods
sense of "planning"; the necessary relationship to derive from techniques for nonlinear
computer programming was incidental to the programming that were developed and
choice of name [5]. Hence the phrase "LP popularized in the 1960s by Fiacco and
program" to refer to a piece of software is not a McCormick, but their application to linear
redundancy, although I tend to use the term programming dates back only to Karmarkar's
"code" instead of "program" to avoid the innovative analysis in 1984[3].
possible ambiguity.
The related problem of integer programming (or
Although all linear programs can be put into the integer linear programming, strictly speaking)
Standard Form, in practice it may not be requires some or all of the variables to take
necessary to do so. For example, although the integer (whole number) values. Integer programs
Standard Form requires all variables to be non- (IPs) often have the advantage of being more
negative, most good LP software allows general realistic than LPs, but the disadvantage of being
bounds l <= x <= u, where l and u are vectors of much harder to solve. The most widely used
known lower and upper bounds. Individual general-purpose techniques for solving IPs use
elements of these bounds vectors can even be the solutions to a series of LPs to manage the
infinity and/or minus-infinity. This allows a search for integer solutions and to prove
variable to be without an explicit upper or lower optimality.
bound, although of course the constraints in the
A-matrix will need to put implied limits on the Linear and integer programming have proved
variable or else the problem may have no finite valuable for modeling many and diverse types of
solution. Similarly, good software allows b1 <= problems in planning, routing, scheduling,
Ax <= b2 for arbitrary b1, b2; the user need not assignment, and design. Industries that make use
hide inequality constraints by the inclusion of of LP and its extensions include transportation,
explicit "slack" variables, nor write Ax >= b1 energy, telecommunications, and manufacturing
and Ax <= b2 as two separate constraints. of many kinds [2].
Two families of solution techniques are in wide The variables in the LP or IP model of the
use today. Both visits a progressively improving transportation problem will hold the values for
series of trial solutions, until a solution is the number of units shipped from one source to a
reached that satisfies the conditions for an destination. A variable with double subscripts is
optimum. Simplex methods, introduced by used for this problem.
Dantzig [3] about 50 years ago, visit "basic"
solutions computed by fixing enough of the Xij = Number of units shipped from Source i to
variables at their bounds to reduce the constraints Destination j
Ax = b to a square system, which can be solved
for unique values of the remaining variables.
Proceedings of the Postgraduate Annual Research Seminar 2005 143
Xij ≥ 0 for all i and j The cost that was minimized in the integer
programming model is far more minimized than
the linear programming model but this shows
6.0 Mathematical Model that in real scenario integer programming model
have changed the input data to an integer format
The Linear Programming model that has been and produces the answer that is far more
used for this transportation problem is as minimized than the answer that was produced
follows: from linear programming model. This shows that
in real world scenario, integer programming
Minimize z = 9.4590 x11 + 16.4415 x12 + produces an outcome that has vast different than
28.3995 x13 + 14.9440 x21 + 29.2110 x22 + the real optimized answer that was produced by
19.1010x23 the linear programming method. Therefore,
Subject to x11 + x12 + x13 ≤ 103.4450, linear programming should be used for
x21 + x22 + x23 ≤ 197.3350, projecting a real world optimization because
x11 + x21 ≥ 71.3500, when considering the outcomes of both of the
x12 + x22 ≥ 133.4980, models, the different are 101.73 Hundred Million
x13 + x23 ≥ 96.1005, Ringgit is very big to taken into account.
x11, x12, x13, x21, x22, x23 ≥ 0.
9.0 Acknowledgements
The Integer Programming model that has been
used for this transportation problem is as This research paper has been studied as a part of
follows: ship routing optimization problem. This research
paper was fully guided, supported and all the
Minimize z = 9x11 + 16x12 + 28x13 + 14x21 + credits should be given to Associate Professor
29x22 + 19x23 Abd. Manan Ahmad. He has always with me to
Subject to x11 + x12 + x13 ≤103,
x21 + x22 + x23 ≤197,
Proceedings of the Postgraduate Annual Research Seminar 2005 144
guide me and advice me to complete the research [3] Blum, C., Roli, A. Metaheuristics in
paper successfully. Thank you very much “PM”. combinatorial optimization: Overview and
conceptual comparison. ACM Computing
References Surveys (CSUR) 35,2 (September 2003),
268-308.
[1] D. O. Casco, B. L. Golden and E. A.
Wasil, “Vehicle routing with backhauls: [4] Winston, W. L. Operations Research:
models, algorithms and case studies,” Applications and Algorithms (Third
Vehicle Routing: Methods and Studies, Edition). International Thompson
Eds: Golden and Assad, North Holland, Publishing, 1994.
Amsterdam (1988).
[5] N. Andreasson, A. Evgrafov, M.
[2] Hamdy, T. Operations Research- An Patriksson, An Introduction to
Introduction (Sixth Edition). Pearson Optimization: Foundations and
Education Inc, 2002. Fundamental Algorithms, Chalmers
University of Technology Press, 2005