A Transportation Problem is a special case of the MCNFP where there is a set S of n supply nodes, at set D of m demand nodes, no transshipment nodes and no upper bounds on arc flows. Because there are no transshipment nodes or upper bounds on arc flow, the MCNFP LP formulation for the transportation problem simplified to
If the total demand is greater than the total supply, then problem is infeasible. If the total demand is equal to the total supply, the problem is said to be a balanced transportation problem. If the total supply is greater than the total demand, we can add a dummy demand node to create a balanced problem.
The lp formulation for a balanced transportation problem is
The constraint matrix for a balanced transportation problem
has a special structure for which researchers have developed more efficient
special-purpose variants of the Simplex method.
GT Railroad has locomotives at IE Junction, Centerville
and Wayover City and needs to get locomotives to the following destinations:
A-Station, Fine Place, Goodville and Somewhere Street. A single locomotive
is required at each destination. Currently there are four locomotives at
IE Junction, one at Centerville, and two at Wayover City. GT Railroad can
send a locomotive from any of the three origin locations (IE Junction,
Centerville and Wayover City) to any of the destinations. For example,
they could send one locomotive to each destination from IE Junction. The
goal, however, is to minimize the total distance that the locomotives are
sent. This may be because the cost of sending a locomotive from one place
to another is directly proportional to the distance traveled. The distances
between the origins and destinations are given in the following table.
A-Station | Fine Place | Goodville | Somewhere Street | |
IE Junction | 13 | 35 | 42 | 9 |
Centerville | 6 | 61 | 18 | 30 |
Wayover City | 15 | 10 | 5 | 9 |
Formulation as a Transportation Problem
Minimize 13 X11 + 35 X12 + 42 X13 + 9 X14 + 6 X21 + 61
X22 + 18 X23 + 30 X24 + 15 X31 + 10 X32 + 5 X33 + 9 X34
Subject to
X11 + X12 + X13 + X14 <= 4 Supply at IE Junction
X21 + X22 + X23 + X24 <= 1 Supply at Centerville
X31 + X32 + X33 + X34 <= 2 Supply at Wayover City
X11 + X21 + X31 >= 1 Demand at A-Station
X12 + X22 + X32 >= 1 Demand at Fine Place
X13 + X23 + X33 >= 1 Demand at Goodville
X14 + X24 + X34 >= 1 Demand at Somewhere Street
Xij >= 0
Formulation as a Balanced Transportation Problem
Since total supply is 7 and the total demand is only 4, we add a dummy demand node with a demand d5 = 3. We also add zero-cost arcs from the supply nodes to the dummy node. Thus, the LP formulation changes to:
Minimize 13 X11 + 35 X12 + 42 X13 + 9 X14 + 6 X21 + 61
X22 + 18 X23 + 30 X24 + 15 X31 + 10 X32 + 5 X33 + 9 X34
Subject to
X11 + X12 + X13 + X14 + X15 = 4 Supply at IE Junction
X21 + X22 + X23 + X24 + X25 = 1 Supply at Centerville
X31 + X32 + X33 + X34 + X35 = 2 Supply at Wayover City
X11 + X21 + X31 = 1 Demand at A-Station
X12 + X22 + X32 = 1 Demand at Fine Place
X13 + X23 + X33 = 1 Demand at Goodville
X14 + X24 + X34 = 1 Demand at Somewhere Street
X15 + X25 + X35 = 3 Demand at Dummy Node
Xij >= 0
Observe that if we add up the supply constraints we get
(S) X11 + X12 + X13 + ? + X33 + X34 + X35 = 7.
If we add up the first four demand constraints, we get
(D) X11 + X21 + X31 + X12 + X22 + X32 + X14 + X24 + X34
= 4.
Subtracting (S) from (D) we get X15 + X25 + X35 = 3. Thus,
the demand constraint for the dummy node is implied by the other constraints
and we can drop it from the formulation. In general, the LP formulation
for a balanced transportation problem has the property that any one constraint
is implied by the remaining constraints.
Theorem: Every BFS for a balanced transportation problem lp is integer-valued.
For many applications, the balanced transportation problem is actually an integer program. As a result of this theorem, however, we can use the Simplex method to solve balanced transportation problems.