GA Router Version 3: Dec 2000 : Fast

This GA uses a swarm of ant finite state automata to search out a path in the two dimensional dual-layer circuit board, each electrically distinct connection is formed by a net of node connections that are attempted by one ant per node.
Average run-time is 0.5 swarm evaluations/sec

Source: RouterAnimApplet4.java Ant4.java
© 2000 F. Michael O'Brien V.2000.12.03

GA Router Version 3: Dec 2000 : Full

This GA is the same as above but with extra diagnostics and statistics for each ant. Since we are only interested in the long-term behavior of this population of swarms of ants competing, use of the faster version above is necessary.
Average run-time is 0.2 swarm evaluations/sec

Source: RouterAnimApplet4.java Ant4.java
© 2000 F. Michael O'Brien V.2000.12.03

GA Router Version 2: Nov 2000

This GA was an initial attempt to provide ant-swarm solution capabilities; but the model was to simplified even a non-optimal solution was not obtainable, a better state model was needed like the version above.

Source: RouterAnimApplet3.java : Ant3.java
© 2000 F. Michael O'Brien V.2000.11.10

GA Router Version 1: Oct 2000

This first GA implementation was a direct mapping of the permutations each cell of the circuit board could obtain. The fitness function used was simply to enumerate all the possible connections between the two devices. This GA approach approached local-maximums too quickly due to the inability of the fitness function in preventing homogeneous populations , where the entire population is dominated by one individual type and only mutation causes further specialization.

Source: RouterAnimApplet.java
© 2000 F. Michael O'Brien V.2000.10.06

See: GA Router Analysis.