This document is based on the NetLogo Programming Guide. It describes how the NetLogo’s programming language is translated in the R language.
NetLogoR, the world is either a
worldMatrix object (similar to a
matrix) or a
worldArray object (similar to an
array which is a collection of matrices). It represents a two dimensional landscape, divided into a grid of square patches (i.e., the matrix cells) of resolution 1.
There are two types of agents in
NetLogoR: the patches and the turtles. Patches cannot move whereas turtles can move. Patches are represented as
matrix objects with two columns
pycor, representing the spatial location (i.e., coordinates) of the center of the patches. See
help("worldMatrix-class") for more details on the patch coordinates system. Turtles are defined by their coordinates
ycor and data associated in
agentMatrix objects. Patches coordinates are always integers but turtles coordinates can have decimals (i.e., a turtle can be positioned anywhere on a patch).
worldArray are similar to
RasterStack in the
raster package. However, there are several differences. Two important differences are the mapping of X and Y coordinates and whether integer locations are in the center or at the corners of integer coordinate systems. In
(0, 0) is at the bottom left, like a mathematical graph, whereas it is in the top left in
Raster* objects, as in geographic coordinates. In
worldArray, integer coordinates are in the middle of the patch or cell, i.e.,
(0, 0) is in the middle of the patch in a
worldMatrix that has 1m resolution and integer coordinates, whereas,
(0, 0) would be at the edge of a cell in a
Raster* that has 1 m resolution and integer coordinates.
agentMatrix is similar to
sp package, with a few differences. The key difference in this case is for speed.
agentMatrix is based on matrices and has no coordinate reference system, both of which make operations dramatically faster than with
SpatialPointsDataFrame objects. In all cases of similar object types, we have implemented functions to convert between the types.
Links are not implemented in
NetLogoR, functions act on the worlds and on the agents. Functions can modify or used them to compute some values. Functions available from the R software and from different packages can be used, as well as the functions from
NetLogoR. Most of the
NetLogoR functions are a translation into the R language of the NetLogo primitives (i.e., “built-in commands and reporters”). You can create new functions or wrap several existing functions into one. The rules to do so are imposed by the R language.
R functions take arguments which are R objects or values. Arguments are used to carry out actions and compute results.
# Create a world according to a given extent w1 <- createWorld(minPxcor = 0, maxPxcor = 10, minPycor = 0, maxPycor = 10) # Report the distance between the patch [pxcor = 0, pycor = 0] and the patch [pxcor = 1, pycor = 1] pDist <- NLdist(agents = cbind(pxcor = 0, pycor = 0), agents2 = cbind(pxcor = 1, pycor = 1, world = w1, torus = TRUE)
NetLogoR, the function arguments indicate which world and/or agents (i.e., patches and/or turtles) are involved in the function. For example, in the
NLdist function, the argument
agents takes one patch coordinates, the argument
agents2 takes another patch coordinates, the argument
world takes a
worldMatrix or a
worldArray object to indicate in which world
agents2 are located and the argument
torus indicate that the world is wrapped.
The result/output of a function needs to be assigned to an object to be kept and re-used later in the model. Functions in R create new objects, they do not directly modify the ones given as arguments. For example, when moving turtles around using the
fd function, the output of the function (i.e., the turtles with new coordinates) needs to be re-assigned to the turtles object.
In NetLogo, pressing a button runs some code. To perform the button action in
NetLogoR, the code must be sent to the R console to be executed. The code can be sent directly by the observer or by a scheduler function (e.g., see
SpaDES package) which executes the code according to the schedule.
The R language uses
# to add comments to the script. Comments make your code easier to read and understand, but they don’t affect its behaviour.
R objects that are neither worlds, patches or turtles can be seen as global variables.
To create a global variable, you create an R object. e.g.,
score <- 15.
Agent variables are places to store different values for an agent. An agent variable can be a patch or a turtle variable. A
worldMatrix object holds only one variable (i.e., value) per patch. A
worldArray object holds several variables per patch (i.e., from the different
worldMatrix stacked). Turtles variables are stored as columns in the data of the
agentMatrix object that represent them. You can have patch and turtle variables defined by the same name as they are not stored in the same object.
To create a new patch variable, you create a
worldMatrix and assign values. The name of the object to which the world is assigned is treated as the name of the patch variable.
Several turtle variables built-in in NetLogo are created at the same time when creating a turtle object in
color. New turtle variables can be created:
Turtle’s breed is a turtle variable and you can have different breeds behaving differently. For example, you could have breeds called
wolf, and have the wolves trying to eat the sheep. You can either specify different breeds in the same turtle object or create two objects, one sheep and one wolf.
You define turtles’ breed by filling the breed column when creating the turtles.
Or you can modify the turtles’ breed after creation.
Global variables are available for all agents. Patch and turtle variables are available to other agents if they can be identified by their coordinates or some other variables.
Local variables may occur in
NetLogoR as inherent to the R software. For example, variables only defined inside functions are local variables and therefore cannot be accessed outside of the functions.
An agentset is a set of agents and you can construct agentsets that contain only some patches or some turtles. An agentset cannot contain the two agent types (patches and turtles) at once. A patch agentset is a
matrix that contains patches coordinates; a turtle agentset is an
agentMatrix object containing turtles. Agentsets can then be passed on as arguments in functions. Agentsets can be redefined at any time.
NetLogoR functions act as if the world is bounded and does not “wrap”. Patches on the sides of the world will have fewer than 8 neighbors and turtles will not move beyond the edges of the world. What happen to the turtles when they reach the edge of the world must be defined.
However you can make the world a torus (“wrap”), so that each patch has the same number of neighbor patches (some located on the other side of the world) and when a turtle moves past the edge of the world, it disappears and reappears on the opposite edge. The option for a torus world is an argument in some of the
NetLogoR functions and can be set
It is common that time passes in discrete steps. Time can be defined as a global variable in the model setup (e.g.,
time <- 0) and then can be incremented in the model when needed (e.g.,
time <- time + 1 at the end of the procedure affecting all the agents).
If a for-loop or a scheduler function (e.g., see
SpaDES package) is used in the model, time does not have to be explicitely defined with a global variable and is handled internally.
Lists may be seen as global variables holding one or more values and strings may be seen as global variables holding characters. Lists and strings are inherent to the R language. Please see R manuals for any related questions.
NetLogo uses lists to store elements. In R, it is faster to use vectors when possible.
R is a software dedicated for data management and statistical analyses. The notation and basic rules of math, as well as the random number generation, are inherent to the R software. Please see R manuals for any related questions.
Reading and writing files is inherent to the R language. Please see R manuals for any related questions on basic functions, as well as the different packages, for more complex functions to manage files.
In R, a file is loaded/opened by assigning its content to an R object. The R object can be read, used and modified but this does not affect the original file. The modified R object can later be written out as a file, either by overwriting the original file or by creating a new file.
There is no interface built-in as part of
However, you can use different plot functions for visualizations similar to the “View” of NetLogo.
plot(nameWorld) works with both
worldArray. To visualize only one layer of a
plot(nameWorldArray[["layerName"]]). Turtles can be plotted alone with
plot(nameTurtles) or on a world already plotted with
Plot() functions (with a capital “P”) from
quickPlot give a better rendering when plotting multiple time steps. These functions work the same way as the
plot() ones, except when adding turtles on a world, use
Plot(nameTurtles, addTo = "nameWorld") instead of
Users can also create any other figures (e.g., the number of turtles through time) on the R plot interface or generate outputs directly on the screen (e.g., show the current time step).
All these procedures need to be coded in the model, either using the
Plot() functions adapted to the
NetLogoR classes for the world visualization, or using R functions and of different packages for any other figures. Please refer to their manuals for any related questions.
The following steps describe how to build a basic NetLogo-type model in R using
NetLogoR. Examples of R scripts are available in the “examples” folder, which can be found using:
Open a new R script (using RStudio or an adapted text editor like Tinn-R) to write the code of the model in it. Save it under the name of the model in an appropriate folder.
Start by naming and describing the model using the symbol
# in front of each line to “comment” the information. This description section is not mandatory but very helpful. It can be considered like the “Info” tab in NetLogo but, in a R script, this section should be kept short and concise.
Load the needed packages (e.g.,
NetLogoR) with the function
library(). This can be done at any time in the code section but must appear before using any function from a package. It is convenient to load all the packages needed at the beginning.
Define the global variables as R objects. Global variables can be defined at any time in the code but it must be done before they are used. Buttons in NetLogo used to set variables must be defined as global variables at the beginning of the R script.
Create the world in which the agents will evolve with the function
createWorld(). Defining the world’s extent in
createWorld() is similar as going in the “Interface” tab in NetLogo and clicking on the “Settings” button to change the model settings.
Define and assign values to the patches when creating the world
createWorld(..., data = ...) or after using the function
NLset(). You can visualize the world with
Create the turtles (i.e., moving agents) with the function
createTurtles(). You can visualize the turtles by plotting them on the world with
points(nameTurtles, pch = 16).
Create the different procedures (i.e., functions affecting the agents) by using the
NetLogoR functions, the R functions or some from other packages.
It is useful to test the different procedures individually with a small world and a few numbers of turtles to make sure the code is doing what you want it to do.
Then, build the main procedure for the model. A for-loop or a scheduler function (e.g., see
SpaDES package) can be used to manage time. The functions placed inside the for-loop or the scheduler function will be iterated the number of time steps defined. The for-loop automatically increases the time step and the scheduler function internally manages time so no “tick” variable is necessary. Executing the for-loop or the scheduler function correponds in NetLogo as hitting the “Go” button in a forever mode. The visualization can be updated at each time step and/or plotted at the end when the iterations are over. Remember that plot functions take time to be executed and can slow down the model.