• Kazan Glenn
• kazan@kazan.co.uk
• written Summer 1995

This is a old program written in 1994/1995. It is interesting to see what could be done with very low specification machines! My original text follows.

This program simulates an environment with a population of 'bugs' within it. The bugs are not static entities but go through changes in response to the environment and also other bug species that may be present. One species can go through drastic alteration and produce further species, when the situation is right, somewhat akin to adaptive radiation. Population fluxes will be observed as bugs interact with their food supply. The user can interact to some degree with this process by observing graphs of population, number of species and also lineage of the species present. Saving and recalling runs is also possible, although it is up to the user to manage the number of files they have stored. See file conventions. Interactions can be carried out either through the keyboard or using the mouse.

Bugs was written using the Zortech C++ v2.1 package on a 486 sx25 machine under MS-DOS with only 4mb of RAM. It is intended to run on almost anything, although it might be rather slow on lower specification machines. It is not intended for use under windows. It is not recommended for use with a black and white monitor as it relies heavily on colours for representational purposes. Use of low resolution graphics for the main window is for two reasons. First it allows the graphics to be faster, and fast graphics was one of the goals of this program. Second, I did not have access to a high resolution graphics mode that could give me the number of colours required.

Little needs to be said except that it is worth noting that a given saved run is actually two files. Theses two files are both binaries, one contains the information regarding the 'bugs' and the state of the world, the other contains all the run information such as population. When loaded this file is constantly open and written to. If you delete one you must delete the other to avoid potential problems. *.WOR files are the world files, *.STT are the statistics and graphing files.

The interface to the file management system as it stands is still crude. No provision has been made for very large numbers of files. It is up to the user to manage their saved runs from the DOS environment. Later versions will contain a rewrite of this interface.

When working from within the program only the run name shared between both file types is given. You do not have to deal with the separate files. Do not add file name extensions. The name 'TMP' is reserved. Do not save any runs under this name.

Intro

The program starts with an introduction screen. Pressing any key will move you on to the primary menu system with three options; starting a new run, loading a saved run or exiting. Do not attemt to load a run unless there are runs present on your disk.

Main

This is the screen displaying the run in progress. While in this screen the mouse driver is deactivated for speed. Pressing any key will return you to the main menu.

Menu

The main menu expands once a run is started to allow you to save, view the networks or view the population trends as autoscaling graphs.

Statistics information

As stated the graphs displaying population trends are autoscaling. As yet there is no way to zoom in to view a small period of time in a long run. Future versions will have this modification. Samples stored in the *.STT files are recorded every 100 cycles. A cycle is defined as being the movement of all existing bugs once.

Viewing networks

The networks are displayed as a graph. Pictographic representation would not have been very clear owing to the large number of connections and the specific nature of these connections. The colour of the square representing a connection indicates the action. A colour coding chart is displayed next to the graph. The number in black over this colour indicates the state to which the connection goes, and therefor the state from which the next action will be derived.

Saving and loading a Run

To save or load a run go to the main menu and press the appropriate button with the mouse or key letter from the key board. Simply enter the name you have chosen, a maximum of eight characters long, with no extensions, and press return. After you will be returned to the main menu.

The Basics

The 'bugs' of this program are actually relatively simple finite state machines. A finite state machine is a relatively old concept derived from cybernetics and the desire to explain and model homeostatic systems. The principle points to be made start with the fact that they are serial in action, unlike a neural network which operates in parallel (within a given layer and dependent on architecture). At any given point they are in a given state or cell. This cell receives the input from the ennvironment, whatever it may be, and produces to things in consequence. The first is an output signal, an action. The second is a state or cell change. This is a simplistic explanation and of course there are many ways to represent these structures and many variations, such as the probabilistic networks used in speech recognition known as hidden Markov models. The inputs and active outputs can be viewed as alphabets. In the simplest cases these both contain two possibilites. In other words they are binary either producing 1's or 0's. However there are no limits and no need for the two sets to contain the same sized alphabets.

Bug Network Specifications

The alpahabet of the bug nets here are different. There are seven possibilities in the input alphabet: food, wall, nothing, same species of bug, different species of bug, same pheromone or different pheromone. There are ten possibilites in the output alphabet: move ahead, move left, move right, these three again but also leaving a pheromone unit behind, turning left but staying in the same place, turning right, remaining still and finally reproducing.

Bugs reproduce under two conditions. First, they reach their maximum energy levels from food consumption, as indicated on the right hand side of the energy indicator while viewing nets. Secondly, akin to a cancer gene, they can have the instruction to reproduce within their networks. This latter case can produce dramatic population explosions, but is a rare gene to aquire through mutation. The maximum energy of a bug is as prone to mutation as the net during the process of reproduction.

The Environment

Food Units

Food units have their own rules governing growth rate and pattern. Principally they spread from covered areas but there is also a degree of random reseeding. This is necessary to ensure a continued 'plant' population. These plants are the dark greeny blue symbols on the main window.

Walls

Walls are the shaded grey blocks. Bugs can not move through or over these objects. They must move around them.

Pheromones

Pheromones are the coloured dots on the display. The colour of the dot is an indication of the species it is derived from. Any given bug determines whether a pheromone unit comes from the same species as itself or from another species. This information is then fed into the network.

Bugs

These are the fast moving things running around the screen! Colour is an indication of species. Bugs will eat other species if they land on the same square, but will not eat their own kind. The pheromone colour they leave (if at all) is the same as their central pixel colour.

Alternative Environmental Situations

There are several possible internal wall formats to chose from. These vary in complexity and therefor hardness. The harder ones are more likely to result in extinction. The best advice here is experiment and observe. If a seeding fails first time try again. No two runs are ever the same, even in the same environment

When bugs reproduce they are subject to mutations. Small mutations of instruction or redirection of the next cell or maximum energy are contained within one species. Large changes such as the addition or removal of a state result in the production of a new species. Thus, within this simulation, there is intraspecific and interspecific competition. This, coupled with the 'bottle necking', 'adaptive radiations' and environmental demands create a plausible if simple model of the process of evolution.

This is a simple and brief explanation of the software. For further information you can contact me at the address given at the top of this file or at the exit window of the program. Thank you for trying this program and I hope it provided some amusement. Feedback is welcome.

Kazan Glenn.