Title: Facility location
1Facility location
2Service facility location
- Geographic representation
- Location problems are based on how geography is
modeled. - In 2D, location can be modeled as xy coordinate
plane, or in a network. - Two ways of measuring distances between two
points with coordinates (xi, yi) and (xj, yj) - Euclidean distance
- Metropolitan metric distance
3Service facility location
- Number of facilities
- Single facility location problem can be solved
analytically with relative ease. - Multi-facility location problems are difficult
because of demand assignment at nodes. Also for
some services, the type of facilities may vary. - Optimization criteria
- Objective function in the location problem
different for different services. Depending on
the owner of the facility, the objectives could
be - Maximization of utility, profit, social benefit
Minimization of travel times, cost.
4Single dimension single facility location problem
- Suppose we wish to find location for a single bus
stop that will serve all the boys-hostels from
Cauvery to Mandakini. - Fix a arbitrary point on the road as origin.
5Single facility location problem
- Result says that the bus stop should be located
at the median w.r.to demand density. - This method is called cross-median approach.
- The result can be extended to 2D using either
form the distance measurement (Euclidean or
Metropolitan).
6Single facility location problem
- Since x and y coordinates are independent of each
other. We can solve two sub-problems (one for
each axis). - Thus the optimal location of the service facility
will have - xs at the median value of wi ordered in
x-direction - ys at the median value of wi ordered in
y-direction - Because each (or both) of the coordinates may be
unique or lie within a range, the optimal
location may be at a point, on a line or within a
region.
7Example
Median weight (7135)/2 8
8Example Solution
9Example Solution
10Example Solution
- Non-unique optimal solution (2,2) and (3,2).
- Hence the line segment joining these two optimal
solution is also an optimal location. Weighted
distance of all the location from any of these
optimal point will be same. Verify!
11Single facility location with Euclidean metric
distance
12Single facility location by center of gravity
method
- An intuitive but non-optimal approach, because it
does not minimize the travel distances from each
location to the service facility. - Center of gravity solution could serve as a
starting point for the previous, rather tedious,
method.
13Single facility location by center of gravity
method
- The optimal solution of the Euclidean method will
always result in a point and will rarely match
the center of gravity solution. - Example of Euclidean method for location problem
FedEx
14Locating a retail outlet
- Objective profit maximization
- Countably finite number of alternative locations
evaluated to find the most profitable site. - Method developed by David L. Huff.
- Basic idea Attractiveness of a facility is
directly proportional to its size and inversely
proportional to the distance.
15Locating a retail outlet
Shopping center may have the parameter value 2
whereas for regular grocery store, this could be
10.
16Locating a retail outlet
- The probability that the customers might be
attracted to other stores is captured by
calculating such probability - Now we can calculate the total consumer
expenditure for product class k at each facility
j.
17Locating a retail outlet
18Example
19Example
- From the previous discussion, assume that a
facility is located at optimal point (2,2).
Assume that we want to check profitability of
location (3,2). - Assume
- Ci weight100.
- Each customer order represents an expenditure of
approximately Rs 10 and ? 2. - The new location at (3,2) is twice as big as the
current one at (2,2). - Travel distances are given in the next table
20Example
21Example
22Example
23Example
Proposed site will have higher market share than
the current location. Hence, the proposed
location is better.
24Multiple facilities Location set covering problem
2
9
20
40
40
30
1
30
7
20
35
20
8
30
30
3
15
30
4
25
6
15
Site within 30 miles could be served by a node.
15
5
25Multiple facilities Location set covering problem
26Maximal covering problem
- The set covering problem minimizes the number
of facilities located under the constraint that
all demand locations need to be served. - However, we have not considered budgetary
constraints. - These budgetary constraints might not allow us to
serve all demand locations. - In these cases, the problem is to locate the
budgeted facilities to maximize the serviced
demand. - These problems are called maximal covering
location problems. - On the other hand, when the objective is to
minimize the total weighted distance traveled
from all demand centers to the opened facilities,
the problem is called a p-median problem.
27Maximal covering problem
- In addition to maximizing the serviced
population, additional constraints may be imposed
within the maximal covering problem. - For example, for emergency facility locations, it
may be desirable to locate facilities in order to
maximize the population which can be served
within five minutes while insuring the entire
population can be served within ten minutes. - This secondary or reduced service time
requirement is included by means of mandatory
closeness constraints.
28Maximal covering problem
- Let there be J 1,2,n demand centers with known
locations and demands pj. In addition, let there
be I 1,2,m possible sites, where a maximum of
K facilities may be opened. - Let the maximum allowable response time (or
distance) be R, which is specified. - If the minimum of facilities required to serve
all the demand centers within the given response
time (or distance) exceeds the allotted (K),
then not all demand centers may be served and we
attempt to serve the maximum population. - Let tij be time (or distance) from ith site to
jth demand center. - The problem formulation then becomes
29Maximal covering problem
- For pj 1 for all js, the problem is same as
the set covering problem and will minimize the
of facilities while serving all demand centers.
For demand distribution not uniform, the optimal
set of facilities will serve maximum population. - Mandatory closeness constraints The demand
centers be located no more than T time units (or
distance) from an open facility, where T gt R. - We have to include these additional mandatory
closeness constraints in our problem.
30Multiple facilities Bowman-Stewart formula
31Multiple facilities Bowman-Stewart formula