Skip to content

Development repository for the Digital Terraria Lab implementation of the Sugarscape agent-based societal simulation.

License

Notifications You must be signed in to change notification settings

nkremerh/sugarscape

Repository files navigation

SUGARSCAPE

Python implementation of the Sugarscape agent-based computational model introduced in Growing Artificial Societies (1996) by Epstein and Axtell.

Inspired by previous work from Herve Lange (https://github.com/langerv/sugarscape) and Joshua Palicka (https://github.com/joshuapalicka/sugarscape).

Requirements:
Python 3

Optional Dependencies:
Matplotlib
TkInter

Usage:
python sugarscape.py

Makefile Options:
make clean
    Clean up working files and logs created by the software.
    Note: This will remove any JSON files created by the other make options.

make data
    Run a number of random seeds comparing selected decision models.
    Note: Results will be saved in the data subdirectory.

make plots
    Generate graph plots from any JSON files in the data subdirectory.
    Note: Plots are dependent on a dataset existing and will create it if necessary.

make seeds
    Generate a set of random seeds comparing selected decision models.
    Note: Results will be saved in the data subdirectory.

make setup
    Change preconfigured settings for the system Python alias.
    Note: Changed settings may alter Makefile and configuration file in-place.

make test
    Run the simulation using the default config.json file and storing a local log in the log.json file.

Preconfigured Examples from Growing Artificial Societies:
A selection of examples can be found in the examples directory.
Each demonstrates a concept from the book Growing Artificial Societies.
Examples are included to demonstrate adherence to or deviation from the source material.

immediate_growback.json:
    Agent sugar collection with immediate sugar growback (pgs. 21-26).

constant_growback.json:
    Agent sugar collection with constant sugar growback (pgs. 28-30).

agent_replacement.json:
    Agent replacement with constant sugar growback (pgs. 32-33).

seasonal_migration.json:
    Agent migration with seasonal sugar growback (pgs. 44-46).

pollution_formation.json:
    Agent sugar collection with constant sugar growback and pollution (pgs. 45-50).

reproduction_basic.json:
    Agent reproduction with constant sugar growback (pgs. 55-58).

reproduction_oscillation_small.json:
    Agent reproduction with lower infertility age and constant sugar growback (pg. 64).

reproduction_oscillation_large.json:
    Agent reproduction with lower reproduction cost and constant sugar growback (pg. 65).

reproduction_oscillation_severe.json:
    Agent reproduction with lower infertility age, lower reproduction cost, and constant sugar growback (pg. 66).

reproduction_inheritance.json:
    Agent reproduction with child wealth inheritance and constant sugar growback (pgs. 67-68).

cultural_tagging.json
    Agent cultural tagging with constant sugar growback (pgs. 72-79).

combat_unlimited.json:
    Agent combat with unlimited combat loot and constant sugar growback (pgs. 82-83).

combat_limited.json:
    Agent combat with unlimited combat loot and constant sugar growback (pgs. 86-90).

cultural_combat_unlimited.json:
    Agent combat with agent cultural tagging and unlimited combat loot (pg. 91).

spice_growback.json:
    Agent sugar and spice collection with constant sugar and spice growback (pgs. 96-99).

trade_basic.json:
    Agent sugar and spice collection with trading and constant sugar and spice growback (pgs. 101-107).

trade_replacement.json:
    Agent trading with agent replacement and constant sugar and spice growback (pgs. 120-122).

trade_reproduction.json:
    Agent trading with agent reproduction and constant sugar and spice growback (pg. 124).

trade_tagging.json:
    Agent trading with agent cultural tagging and constant sugar and spice growback (pgs. 125-126).

trade_tagging_reproduction.json:
    Agent trading with agent cultural tagging, agent reproduction, and constant sugar and spice growback (pg. 127).

trade_pollution.json:
    Agent trading with sugar pollution and constant sugar and spice growback (pgs. 127-129).

foresight_basic.json:
    Agent sugar and spice collection with foresight consideration and constant sugar and spice growback (pgs. 129-130).

lending_basic.json:
    Agent sugar and spice collection with lending and constant sugar and spice growback (pgs. 131-133).

disease_basic.json:
    Agent sugar and spice collection with disease and constant sugar and spice growback (pgs. 141-147).

JSON Configuration File Options:
The simulation provides a default set of options in a dictionary in the sugarscape.py file.
A JSON configuration file can be passed to the simulation, overwriting the default configuration, with the --conf option.

agentAggressionFactor: [float, float]
    Set the aggressiveness of an agent.
    Note: The more aggressive an agent the more likely they will be enticed by combat options.
    Default: [0, 0]

agentBaseInterestRate: [float, float]
    Set the interest rate for an agent's lending as a percentage.
    Default: [0.0, 0.0]

agentDecisionModelLookaheadDiscount: [float, float]
    Set the agent's discount applied to considering possible rewards in future timesteps.
    Note: This feature is used in agent decision models besides the "none" and "rawSugarscape" models.
    Default: [0, 0]

agentDecisionModelLookaheadFactor: float
    Set the agent's consideration of future rewards out to the provided forecasting horizon.
    Options: 0, 0.5
    Note: This feature is used in agent decision models besides the "none" and "rawSugarscape" models.
    Default: 0

agentDecisionModelFactor: [float, float]
    Set the agent weight of their decision model over biological imperatives.
    Default: [0, 0]

agentDecisionModels: [string, ...]
    Set the agent decision models for different decisionmaking.
    Options: "altruistBinary", "altruistTop", "benthamBinary", "benthamTop", "egoistBinary", "egoistTop", "negativeBentham", "none", "rawSugarscape"
    Note: Adding either "HalfLookahead" or "NoLookahead" after a decision model name will enforce a particular decision model lookahead factor.
    Default: ["none"]

agentDecisionModelTribalFactor: [float, float]
    Set how much agents prioritize the welfare of their own tribe members over others.
    Note: Valid range is [0.0, 1.0], use -1 to disable.
    Default: [-1, -1]

agentDepressionPercentage: float
    Set the percentage chance an agent will experience depression symptoms at birth.
    Note: The starting agent population will have this same percentage of depressed agents.
    Default: 0.0

agentFemaleInfertilityAge: [int, int]
    Set the timestep age at which female agents become infertile.
    Default: [0, 0]

agentFemaleFertilityAge: [int, int]
    Set the timestep age at which female agents become fertile.
    Default: [0, 0]

agentFertilityFactor: [float, float]
    Set the fertility bonus for the agent.
    The higher the factor, the fewer resources the agent expends to reproduce.
    Default: [0, 0]

agentImmuneSystemLength: int
    Set the length of agent immune system tags to integer length.
    Default: 0

agentInheritancePolicy: string
    Set wealth inheritance policy on agent death to given string.
    Options: "children". "daughters", "friends", "none", "sons"
    Default: "none"

agentLendingFactor: [float, float]
    Set lending aggressiveness of agent.
    The more aggressive an agent is to lend, the higher the offered interest rate will be.
    Default: [0, 0]

agentLoanDuration: [int, int]
    Set the agent's provided loan duration in timesteps.
    Default: [0, 0]

agentLookaheadFactor: [int, int]
    Set the agent's consideration of metabolic costs in timesteps ahead.
    Default: [0, 0]

agentMaleInfertilityAge: [int, int]
    Set the timestep age at which male agents become infertile.
    Default: [0, 0]

agentMaleFertilityAge: [int, int]
    Set the timestep age at which male agents become fertile.
    Default: [0, 0].

agentMaleToFemaleRatio: float
    Set the ratio of males to females in starting population as a real number.
    Default: 1.0

agentMaxAge: [int, int]
    Set maximum agent age in timesteps.
    Note: A value of -1 indicates an infinitely lived agent.
    Default: [-1, -1]

agentMaxFriends: [int, int]
    Set maximum number of friends tracked by an agent.
    Default: [0, 0]

agentMovement: [int, int]
    Set maximum movement distance per timestep for agent.
    Default: [1, 6]

agentMovementMode: string
    Set the directionality used to figure out which cells an agent can move to.
    Options: "cardinal", "radial"
    Default: "cardinal"

agentReplacements: int
    Set maximum number of agents to replace in environment on agent death.
    Default: 0

agentSelfishnessFactor: [float, float]
    Set the weight agents provide to their own welfare and to other agents.
    Note: Valid range is [0.0, 1.0], use -1 to disable.
    Default: [-1, -1]

agentSpiceMetabolism: [float, float]
    Set agent metabolism for spice per timestep.
    Default: [0, 0]

agentStartingSpice: [float, float]
    Set agent starting spice hold.
    Default: [0, 0]

agentStartingSugar: [float, float]
    Set agent starting sugar hold.
    Default: [10, 40]

agentSugarMetabolism: [float, float]
    Set agent metabolism for sugar per timestep.
    Default: [1, 4]

agentTagPreferences: bool
    Set whether agents calculate welfare based on cultural tag preferences.
    Note: Requires agents have a cultural tag string length greater than zero.
    Default: false

agentTagStringLength: int
    Set agent cultural tags string length.
    Default: 0

agentTagging: boolean
    Set whether agents will exert cultural pressure by flipping neighboring agent cultural tags.
    Default: true

agentTradeFactor: [float, float]
    Set agent trade aggressiveness.
    Note: The more aggressive in trading an agent, the more resources they will attempt to trade.
    Default: [0, 0]

agentUniversalSpice: [float, float]
    Set the amount agents recieve for universal basic spice income.
    Default: [0, 0]

agentUniversalSugar: [float, float]
    Set the amount agents recieve for universal basic sugar income. 
    Default: [0, 0]

agentVision: [int, int]
    Set the distance in the four cardinal directions an agent can see.
    Default: [1, 6]

agentVisionMode: string
    Set the directionality used to figure out which cells an agent can see.
    Options: "cardinal", "radial"
    Default: "cardinal"

debugMode: [string, ...]
    Set the debug printing mode.
    Options: "agent", "all", "cell", "disease", "environment", "ethics", "none",  "sugarscape"
    Default: "none"
    Note: Some options may cause no output in current impementation.
          Can select multiple debug modes simultaneously.

diseaseAggressionPenalty: [float, float]
    Set the impact a disease will have on an agent's aggressiveness.
    Note: Negative values constitute an aggressiveness decrease.
    Default: [0, 0]

diseaseFertilityPenalty: [float, float]
    Set the impact a disease will have on an agent's fertility.
    Note: Negative values constitute a fertility decrease.
    Default: [0, 0]

diseaseMovementPenalty: [int, int]
    Set the impact a disease will have on an agent's movement distance.
    Note: Negative values constitute a decrease in movement range.
    Default: [0, 0]

diseaseSpiceMetabolismPenalty: [float, float]
    Set the impact a disease will have on an agent's spice metabolism rate.
    Note: Negative values constitute a decrease in agent spice metabolism.
    Default: [0, 0]

diseaseSugarMetabolismPenalty: [float, float]
    Set the impact a disease will have on an agent's sugar metabolism rate.
    Note: Negative values constitute a decrease in agent sugar metabolism.
    Default: [0, 0]

diseaseTagStringLength: [int, int]
    Set the length of disease tags.
    The longer the length, the longer an agent will have the disease.
    Default: [0, 0]

diseaseVisionPenalty: [int, int]
    Set the impact a disease will have on an agent's vision.
    Note: Negative values constitute a decrease in agent vision.
    Default: [0, 0]

environmentEquator: int
    Set the equator of the environment for seasonal changes.
    Note: Value of -1 causes equator to be set at the midpoint of the environment.
    Default: -1

environmentHeight: int
    Set the height in cells of the Sugarscape environment.
    Default: 50

environmentMaxCombatLoot: float
    Set the maximum reward agents receive from winning combat.
    Default: 0

environmentMaxSpice: int
    Set the maximum amount of spice at any cell in the environment.
    This amount will only be present at spice peaks.
    Default: 0

environmentMaxSugar: int
    Set the maximum amount of sugar at any cell in the environment.
    This amount will only be present at sugar peaks.
    Default: 4

environmentMaxTribes: int
    Set the maximum number of tribes in the starting population.
    Default: 0

environmentPollutionDiffusionDelay: int
    Set the delay interval in timesteps when pollution is diffused across the environment.
    Default: 0

environmentPollutionDiffusionTimeframe: [int, int]
    Set the start and end timesteps during which pollution diffusion is active.
    Note: Value of -1 for the start sets the start timestep to 0.
    Note: Value of -1 for the end sets the end timestep to the end of the simulation.
    Default: [0, 0]

environmentPollutionTimeframe: [int, int]
    Set the start and end timesteps during which consumption and production pollution are active.
    Note: Value of -1 for the start sets the start timestep to 0.
    Note: Value of -1 for the end sets the end timestep to the end of the simulation.
    Default: [0, 0]

environmentQuadrantSizeFactor: float
    Set the proportion of each corner of the screen taken up by the agents' starting quadrants.
    Default: 1

environmentSeasonalGrowbackDelay: int
    Set the delay interval in timesteps when resources are regrown when cell is in a dry season.
    Default: 0

environmentSeasonInterval: int
    Set the interval in timesteps when environment seasons change.
    Seasons change along the equator of the environment.
    Default: 0

environmentSpiceConsumptionPollutionFactor: float
    Set the amount of pollution generated by an agent consuming spice at a cell.
    Default: 0

environmentSpicePeaks: [[int, int], ...]
    Set the coordinates for spice peaks in the environment.
    Default: [[15, 15], [35, 35]]

environmentSpiceProductionPollutionFactor: float
    Set the amount of pollution generated by an agent collecting spice at a cell.
    Default: 0

environmentSpiceRegrowRate: int
    Set the amount of spice regrown across the environment per timestep.
    Each cell can only grow up to their maximum spice value.
    Default: 0

environmentStartingQuadrants: [int (,int, int, int)]
    Set environment quadrants in which agents will initially be placed.
    Quadrant 1 begins in the top left.
    Quadrant 2 begins in the top right.
    Quadrant 3 begins in the bottom right.
    Quadrant 4 begins in the bottom left.
    Default: [1, 2, 3, 4]

environmentSugarConsumptionPollutionFactor: float
    Set the amount of pollution generated by an agent consuming sugar at a cell.
    Default: 0

environmentSugarPeaks: [[int, int], ...]
    Set the coordinates for sugar peaks in the environment.
    Default: [[15, 35], [35, 15]]

environmentSugarProductionPollutionFactor: float
    Set the amount of pollution generated by an agent collecting sugar at a cell.
    Default: 0

environmentSugarRegrowRate: int
    Set the amount of sugar regrown across the environment per timestep.
    Each cell can only grow up to their maximum sugar value.
    Default: 1

environmentTribePerQuadrant: bool
    Set whether starting quadrants are initially populated by a single tribe.
    Note: This will overwrite the number of tribes with the number of starting quadrants.
    Default: false

environmentUniversalSpiceIncomeInterval: int
    Set the interval in timesteps when environment produces universal basic spice income.
    Default: 0

environmentUniversalSugarIncomeInterval: int
    Set the interval in timesteps when environment produces universal basic sugar income.
    Default: 0

environmentWidth: int
    Set the width in cells of the Sugarscape environment.
    Default: 50

environmentWraparound: bool
    Set whether the environment is a torus (wraparound) or a plane (no wraparound).
    Default: true

experimentalGroup: string
    Set the experimental group of agents under study for finer-grained logging.
    Options: "depressed", "female", "male", "sick"
    Default: null

headlessMode: bool
    Set whether the GUI is enabled.
    Default: false

interfaceHeight: int
    Set number of pixels for GUI height.
    Note: Values below zero will cause the interface to fit to 1/2 total display height.
    Default: 1000

interfaceWidth: int
    Set number of pixels for GUI width.
    Default: 900
    Note: Values below zero will cause the interface to fit to 1/2 total display width.

logfile: path
    Set the path of the log file.
    Default: null

logfileFormat: string
    Set the file format for the log file.
    Default: "json"

neighborhoodMode: string
    Set the type of neighborhood adjacency used by agents and environment cells.
    Options: "moore", "vonNeumann"
    Default: "vonNeumann"

profileMode: bool
    Set whether performance profiling mode is enabled.
    Default: false

seed: int
    Set the seed value for the random number generator.
    Note: Value of -1 causes simulation to generate a random seed.
    Note: Reusing a seed ensures deterministic simulation outcomes.
    Default: -1

startingAgents: int
    Set the number of agents placed in the initial population.
    Default: 500

startingDiseases: int
    Set the number of distinct diseases at simulation start.
    Default: 0

startingDiseasesPerAgent: [int, int]
    Set the number of diseases given to each agent at simulation start.
    Note: [0, 0] will give each starting disease to a unique agent.
    Default: [0, 0]

timesteps: int
    Set the number of timesteps the simulation runs.
    Note: Value of -1 causes simulation to run forever or until there are no more living agents.
    Default: 200

Other JSON Configurable Options:
decisionModels: [[string, ...], ...]
    Set the agent decision models to be tested in data collection.
    Default: [["none"]]

jobUpdateFrequency: int
    Set the frequency at which the number of remaining jobs is reported.
    Default: 5

numParallelSimJobs: int
    Set the number of simulations to run in parallel during data collection.
    Default: 1

numSeeds: int
    Set the number of random seeds to be tested in data collection.
    Default: 100

plots: [string, ...]
    Set the plots to be created once data has been collected.
    Default: ["deaths", "meanAgeAtDeath", "meanttl", "meanWealth", "population", "wealth"]

plotTimesteps: int
    Set the number of timesteps to plot in graphs as the X axis.
    Note: This option does not control how many timesteps the simulation runs.
    Default: 1000

pythonAlias: string
    Set the alias to the local Python 3 installation.
    Note: Python 3 is required to run the simulation.
    Default: "python"