GENETIC ALGORITHM DEMONSTRATION

by GA Class (CS 495),     Fall 1998,  Instructor: Jeffrey Horn

The NYC Tunnels Problem

a well-known, open problem in pipe network design
 
 

PROBLEM DESCRIPTION:

• supplying clean water to New York City
• why is this problem so hard?  Only 21 pipes, and 16 choices per pipe...

PROBLEM IMPLEMENTATION:  The Pipe Network Simulator

We treat the simulator as a black box that takes a chromosome (representing a pipe network), evaluates it, and returns a total cost, which the GA will try to minimize.

Here is an interface to the pipe network simulator that you can use yourself, to try your hand at designing the NYC pipe network!

pipe simulator

• Motivation:
• The Basic Components:
1. A Population of Chromosomes
2. Then a sequence of "generations"
3. Each generation, apply the basic operators:
4. The Basic Operators:
1. Selection
2. Crossover
3. Mutation
5. To create a new population
6. Evaluate each member of the new population (run through black box)
7. Goto #3 above
8. Repeat until ....
• Advantages of Evolution:
• Parallel Computation (natural, highly parallel)
• Identify and recombine "BUILDING BLOCKS" (i.e., good partial solutions from different individuals)