						----------

						BACKGROUND

						----------



This program was originally coded by Baker in 1991 and extended by Chan and 

Baker in Chan (2005) in Section 4.11 of his book LOCATION, TRANSPORT & LAND-

USE. The study extends the work done by the Military Airlift Command (now the 

Air Mobility Command) Analysis Group on reducing the operating costs of the 

Defense-Courier-Service network.  



The study's primary focus is to minimize those costs by varying the number 

and location of servicing depots, and the routes flown from those depots. The 

theoretical algorithm used in the methodology is an expansion of Laporte's 

(1986) formulation of the multiple-depot multiple-traveling-salesman 

facility-location problem (which is described in the "Measuring spatial 

separation" chapter of Chan [2005].) Multiple servicing-frequency is 

addressed by clustering co-located demands with Kulkarni's (1985) subtour-

breaking constraint. Vehicle range is considered by redressing a shortfall of 

the subtour-breaking constraint, which was first noted by Brodie (1988).  The 

formulation is merely used as a validation of a system-wide solution 

heuristic, since exact solution is beyond the range of current computing 

capabilities. 



The solution heuristic is a combination of the minimum spanning-forest (Prim 

and Dijkstra) and the Clarke-Wright method. The spanning-forest is used for a 

depot location and partitioning of the study area into service regions, while 

the Clarke-Wright algorithm computes the routes flown from the depots to 

their assigned service points. The heuristic is suboptimal by 3.8% on the 

average in six validation runs, with no run greater than 20%. The results 

indicate several depots may be closed without large increase to system 

mileage. The entire case study is documented in the "Simultaneous location-

routing" chapter of Chan (2005) in Section 4.11 as mentioned.



						--------

						SOFTWARE

						--------



The software contained here is organized under the SPANFRST directory. Sample 

data files are included to illustrate program execution. The logic of this 

program is explained in the book under the "Location-Routing" chapter of Chan 

(2005), where the combined location-and-routing model is fully explained. 

This master SPANFRST directory contains three subdirectories.



VALID subdirectory

******************



The subdirectory labeled VALID provides the most comprehensive overview of 

the various files, including other subdirectories such as CW, which is used 

in the validation process. It contains all the files used in model validation 

using the extended Laporte formulation, as mentioned above. Following 

standard practice, there is a readme.txt file within the VALID sub-directory 

which should also be referred to.



MSF subdirectory

****************



The subdirectory labeled MSF contains the FORTRAN file for minimum-spanning-

forest as mentioned above. Since the same code was used repeatedly, only one 

sample input and output file accompanies the code file. Refer to S. F. Baker 

(1991)- LOCATION & ROUTING OF THE DEFENSE COURIER SERVICE - Appendix E for a 

comprehensive listing of various outputs that can be generated.



CW subdirectory

***************



The subdirectory labeled CW contains the FORTRAN file for the Clarke-Wright 

heuristic. The generic algorithm is described in the "Measuring Spatial 

Separation" Chapter of Chan (2005) and the special adaptation here is 

documented in the "Location-routing" chapter of the same book. As before, 

only sample input and output files are included, since the full output is 

given in Appendix E of Baker (1991).



The reader can execute any of these subdirectories independently even though 

they perform an integrated task when they are all invoked. For an 

introduction, the reader is referred to the VALID subdirectory. To the extent 

that this sub-directory contains codes and data that validate heuristic 

against analytic models, it provides the most comprehensive overview of the 

entire software package.

