VRPH is a library of heuristics for quickly generating 
solutions to Vehicle Routing Problems.  VRPH was 
developed by Chris Groer as part of his dissertation at 
the University of Maryland, College Park.  This code is 
provided in hopes that it will be useful to others researching 
computational methods for the VRP.  

To compile VRPH, read the INSTALL file.  There are no
external dependencies and it should be straightforward
to build the code.  Building the code produces at least
4 binaries, all located in the VRPH-1.0/bin directory:

1) vrp_rtr - implementation of a metaheuristic for quickly
generating good solutions to the VRP

2) vrp_sa - implementation of a simulated annealing
algorithm for the VRP.  In general, solutions are not as 
good as those produced by vrp_rtr.

3) vrp_ej - implementation of a neighborhood ejection-
injection algorithm for the VRP.  This can produce solutions
that are better than vrp_rtr and can be used to further
refine these solutions

4) vrp_init - implementation of two classical heuristics
for generating initial feasible solutions

There are several additional binaries that require you to
link with external software:

1) vrp_sp - implementation of a set partitioning-based
algorithm for the VRP.  Many feasible routes are generated
using the metaheuristic in vrp_rtr.  These are added as
columns/variables to the set partitioning problem.
This requires you to link with COIN-OSI and GLPK.  It should
not be hard to use a different MIP solver as there is very
little solver-specific code

2) vrp_plotter - a utility to write graphical illustrations
of solutions to a .ps file.  This requires one to link to 
the open source PLPlot library.

The VRPH-1.0/doc directory is a placeholder for Doxygen-generated
documentation.  See the INSTALL file for details on creating
this documentation via the makefile.

Right now, the code is able to generate solutions to classical 
capacitated VRPs.  Adding support for time windows, multiple 
depots, etc. should not be too difficult and the code was 
designed to make it somewhat easy to add new constraints.  

The various subdirectories contain README's that describe 
the data sets, etc.  See the LICENSE file in this directory for 
details on the licensing of this code.

If you find the code useful, I'd like to know, so send me an email.
If you encounter bugs or problems, I'd like to know that too.

Chris Groer
cgroer@gmail.com
