Godplay

WIP; at this point I'm noting down the plan, with time I'll start implementing

Virtual environment in which fake life tries to thrive.

This is a first iteration of the model, so it describes a world as a plane (its as easy to walk on land and water and there are no hills, so you move everywhere as easy), where some herbivorous creatures live. No specimen ever kills another, but they are pretty dumb and their planning facilities are almost algorithmic ( they are not truly algorithmic, since there is a random factor involved, to simulate irrationality, panic and general absurdity on life at large. Besides, genome model is pretty trivial and birth happens instantly after copulation. These (and some more) are simplifications, but they should be easily extendable and replacable.

Overview

There is a map that has sectors. These sectors have characteristics (some are full of water, some spawn food, some are toxic, warm, cold, safe, etc). In these sectors additional objects (like food) may be also spawned.

On that map there is also an initial population of randomized (well, almost, there should be a pre-run that makes initial population survivable) specimen.

Each specimen has a genotype. That genotype transcribes to fenotype, which is description of that single specimen. It contains its sex, aging process (health, eyesight, speed, etc over time but also expected time of life if dying of old age), daily needs, resiliency to unsatisfied needs, and decision-making process (when replanning occurs, which urges are more important, etc).

Specimen also has state, that contains current position, health, mental state and physical condition (which impacts things like speed or inter-specimen conflict resolution), level of satisfaction of each need (both on daily and long-term basis).

Atomic time intervals are gonna conventionally be called days. It doesn't really work, since in this model specimen doesn't really have to sleep or eat daily, but it's quite easy to use when modelling.

Each specimen has a set of needs (e.g. need for food, sleep, etc). Its fenotype describes daily levels of these needs. If a need won't be satisfied during a day, it accumulates for long-term and long-term needs can be satisfied by providing more than daily need level. Long-term needs are usually "debt", but "surplus" can occur as well (though surplus will gradually decrease with time, even if daily levels are satisfied).

Needs are satisfied by performing actions (eating, sleeping, etc). Actions have requirements (specimen must be in watery area to drink, next to food to eat, etc) and satisfy number of needs at some level (e.g. sleeping in 'normal' place satisfies need for sleep, but sleeping in safe lace satisfies need for sleep better and also satisfies need for safety).

Unsatisfied needs cause urges (not neceserarily any unsatisfied need, more probably these that have debt above some threshold which was described as 'resiliency to unsatisfied needs' above) to take some actions that will satisfy them. Urges have levels based on short- and long-term needs, but other factors may come into play as well (e.g. attractiveness of some other specimen may boost level of sexual urge).

Specimen always have a plan. Plan is supposed to satisfy an urge and consists of series of steps (usually several "move" and "perform action" at the end). From these steps we can derive cost of a plan (how much effort, that loosely translates to time required, is this gonna take?).

On most days specimen will perform next step of current plan. In some cases (I've finished previous plan yesterday, food I wanted was eaten by someone else, my wanted partner died or ran, etc, but also some urges suddenly overwhelmed other urges, together with the one we're trying to satisfy by executing current plan) it may start the day with replanning, which just discards current plan and chooses a new one.

Planning (choosing an action to take and a series of steps to satisfy that action requirements) is based on several factors: current urge levels (how much do I require that?), plan costs (can I have similiar or not much worse effects with less effort?), action results (if I can eat more or less for similiar effort, I should eat more), personal preferences (these map to resiliency to unsatisfied needs; some specimen may have higher need for water than for food, so when both these needs are similarly satisfied, it will probably choose plan that brings water) and some random factor for good luck and some fun.

Details

For the sake of initial values, we assume that average daily need of average specimen over the whole simulation is 10. All the numbers will be tweaked by trial and error once implemented.

World

World is represented as a cyclic n/m grid ((x, 0) is just above (x, m), and (0, y) is just to the right from (n, y)) of cells called region. Each region has any (0-n) number of modifiers:

watery
food-spawning
toxic
dangerous
safe
warm
cold

There are some combinations that cannot go together:

safe region cannot be toxic or dangerous
warm region cannot be cold

and obviously vice-versa; though a region can be simultaneously toxic and dangerous

Evolution

At this point the only thing that changes in the world is that food is spawned.

Future extensions: carnivores, other non-violent species, maybe seasons of the year that tweak regions modifiers or their probabilities.

New food appears in food-spawning regions with some probability. This probability is boosted if a region is additionally watery.

Initial proposed values:

food spawn probability = 1%

watery region boost = 0.5%

Food is described with (gain = how much hunger is satisfies):

initial gain (random from 30-50)
final gain (35-65% of initial gain)
lifespan (random from 60-100 days)

If a specimen consumes food at the same day as that food was spawned, he satisfies his need by initial gain. Every day through the lifespan (measured in days) gain of that food object decreases linearily to final gain (at (spawn time + lifespan) moment, objects gain = final gain). Next day the object disappears.

Future extenion: uneaten food fertilizes ground, increasing its spawn probability.

Specimen

Genetics

Unless explicitly stated, things like daily, value, weight, etc are floats.

Genotype

enum Needs = ...  # see section below
enum Feature = several (lets starts with 10) values without clear semantics (they represent the magic that we're trying to figure out in DNA)
enum Sex = Male, Female

class Genome:
    class Chromosome:
        daily[Need] needs
        value[Feature] features

    # size at least 3 - one male, one female, one general; 
    weight[Chromosome] genes
    # never the same
    Sex sex

    def weightedChromosome() -> Chromosome: # this is more like RNA, final level before transcription to fenotype
        first chromosome is related to being on one sex, last of the other
        when an entity is male, male-related chromose weight is doubled and female-related weight is cut in half
        reverse happens for females

        return chromosome obtained by weighing chromosomes from genes by their weights

Fenotype

class StateOverLife<Unit>: # utility object
    Unit base
    double ageParam

    def value(age: (0, 1]) -> Unit:
        something that simulates aging, with a base value that is multiplied by some function controlled by age param
        (e.g. when reaching half an age, that property may be almost twice the base, while infants and elders are
        down to 75% of base)

class Fenotype:
    Sex sex
    int diesAt
    //this impacts thresholds of satisfactionLevels
    [0.9, 1.1][Need] character
    StateOverLife<Health> healthState # Health is int
    StateOverLife<radius> eyesight
    StateOverLife<radius> speed
    StateOverLife<daily>[Need] urges
    StateOverLife<(0, 0.5]> mentalStrength # controls happiness (psychical condition) threshold
    StateOverLife<(0, 0.5]> physicalResiliency # controls physical condition threshold
    StateOverLife<low-int range> replanningPeriod

We're using term happiness for mental state and (physical) condition to avoid confusion between psychical and physical

Transcription

Mutation

daily mutation breeding mutation

Crossover

details on when it happens - later, at reproduction needs

State

fenotype, age, position, plan, daily and long-term needs saitsfation, timestamp, derivable urges, actions to satisfy needs and to proceed through they day (satisfy() will probably be passed as a callback to action)

FilipMalczak/godplay