First of all, the genetic algorithm on the pFSAs were completely redesigned.
The goal for the majority of the year has been the development of the predator prey simulation. This required a lot of programming and reprogramming. The idea and implementation of different species was necessary, as was the generalisation of actions and observations. CLOS has a very friendly object system, and the actions, observations and much of the environment in general, hadn't taken advantage of this. Now the actions and observations and species are fully modular and object oriented. So a lot of programming and reprogramming and testing was done to allow the predator prey experiments to happen.
The statistics used in the simulation were completely rewritten. Once again full advantage was taken of the Common Lisp Object System; a modular statistics class was designed and implemented. An awk script to extract data from the raw output of the simulation was also written. All the data is output in Gnuplot readable format.
Once the above framework was in place, new actions and observations and statistics needed to be devised. The attack action needed to be written and tested, as did the food selection and food analysis. The body usage statistic and the deliberation measure had to be written, also.
It became apparent that running simulations one at a time was insufficient for these experiments, as a lot of the populations died out under the parameters. To remedy this, the simulation parameters were moved to a simulation running script, and a simple set of tools developed to run experiments on the school's computer cluster.
A number of utilities to assist in data analysis were implemented. These included automatic graphing programs, averaging programs and proportioning programs, all of which read Gnuplot format, and output Gnuplot format.