This package contains basic functions for developing simulated agents that play the game TwoDots. The simulated agents can play the game with the GUI visible, or turned off.

I will provide a simple example in which the agent selects the largest set of dots on each trial. Note that this agent is far from optimal because it does not look ahead to find rectangles, which tend to provide the most points. The stand-alone code associated with this example can be found in models/model_1.

The first step is to install the package with ] add TwoDotsModels or ] dev TwoDotsModels for development.

Load TwoDotsModels and import the function decide! so that it can be extended for this specific model.

using  TwoDotsModels
import TwoDotsModels: decide!

Define a model object that is a subtype of TwoDotsModel. At minimum, the model must contain the field real_time, which runs the model in real time if true.

mutable struct Model <: TwoDotsModel 
    real_time::Bool 
end

Model(;real_time=false) = Model(real_time)

Below, I will define a decide! function for the model. The function search_all returns all connectable sets of dots. In the model, the largest set of connectable dots is assigned to the variable choice. select_choice! is a custom function that iterates through each dot in choice and clicks on it (see below). click_submit is a function that processes the choice, updates the score, and replaces the selected dots with new dots.

function decide!(model::Model, game, gui)
    options = search_all(game)
    n_dots = length.(options)
    _,idx = findmax(n_dots)
    choice = options[idx]
    select_choice!(model, game, gui, choice)
    model.real_time ? sleep(1) : nothing
    click_submit!(game, gui)
    return nothing
end

As described above, select_choice! clicks on each dot in choice followed by a pause if real_time is true.

function select_choice!(model, game, gui, choice)
    for dot in choice
        click_dot!(game, dot, gui)
        model.real_time ? sleep(.5) : nothing
    end
end

The following code block illustrates how to run the simulated agent with the GUI. First, an instance of Model is produced with real_time set to true. Second, an instance of a game is generated with default parameters. Next, an instance of the gui is defined based on the game object. Finally, model, game and gui are passed to the function run!.

model = Model(;real_time = true)
game = Game()
gui = GUI(;game)
run!(model, game, gui)

The simulated agent can be ran without the GUI simply by ommitting the gui object. The default value of real_time is false, which means the simulated game will run as quickly as possible. If you want to run without the GUI in real-time, you can set real_time to true.

model = Model()
game = Game()
run!(model, game)