/****************************************
 *Genetic Algorithm with Wind Farm Application Libraries
 * 
 * Gary Yan, Peter Zhang
 * June 13,2012
 ****************************************/

- Was developed using Armadillo-3.0, but a newer version is available at http://arma.sourceforge.net/download.html



--------------Libraries:
There are three major parts to the code:

- libGA.a
  - contains the functions required for a general GA optimization problem.  GAProblem and GAControl class and member function definitions 
    are included here
  - consists of files:
    -GA.cc
    -src/GAProblem.cc
    -src/GAControl.cc
    
- libWindFarmEvaluation.a
  - contains functions used for the wind farm problem.  Contains the power generation calculation and noise generation for receptor uniformly spaced 40m apart on a 2000m x 2000m square piece of land
  - consists of files:
    -WindFarmEvaluation.cc
    -GAdriver.cc

-GAdriveruser.cc
  - defines parameters in the GA.(in this case the Wind farm optimization)

-----------------Building
- to compile the updated code, type 

  make

on the command line (development used bash) This will compile the source code according to what has changed.
If only GAdriveruser.cc has changed, only that will be compiled.
If any item in the dependency list of libWindFarmEvaluation.a has changed, that and GAdriveruser.cc will be compiled
If any item in the dependency list of libGA.a has changed, that, libWindFarmEvaluation.a, and GAdriveruser.cc will be compiled

type: 

make clean
make

to recompile all sources

-----------------Running
- Once the code is compiled, type

./ga

to run the code.  



-----------------GAdriveruser.cc---------
Defining Parameters
- The application needs to be written by the user in a file, called say GAdriveruser.cc
- There are 2 options for parameter definitions:
  - calling readFile will read inputs in the specified format in a file called files/_input.dat 
  - calling set functions in GAProblem and GAControl class can set parameters as well (see sample GAdriveruser.cc code)


Defining Constraints
One way of defining land constraints is through boolean evaluations.  In the sample GAdriveruser.cc, land constraints were specified using the input coordinate vector x.  A count variable is used to measure the degree of constraint violation by checking if the x-y positions are inside the restricted region.  



- if there are more than 2 objectives, or 2 inequality constraints, or 2 equality constraints, may need to define additional function prototypes in GA.hh




------------------Some Ideas for Future Work
- add other methods of fitness assignment (SPEA, etc) and add to call list in fitnessAssignment function
- add other constraint handling method e.g.:
  - constrained BIN-TOUR, where the normal BIN-TOUR is replaced by selecting:
    - if both are feasible, pick the higher fitness
    - if one is feasible and the other is not, pick the feasible
    - if both are infeasible, pick the one with "lower" constraint violation
- add hypervolume indicator computation for nObj > 2.  This is the largest remaining obstacle for generalization of the GA code for n Objectives.  Other parts handle the general n case well (e.g. the crowding distance computation was checked by hand for n = 3) Many sweeping algorithms are available online that can calculate this in O(n^(d-2) log n ) time