| Title: | A Graph Based Particle Simulator Based on D3-Force |
|---|---|
| Description: | Simulating particle movement in 2D space has many application. The 'particles' package implements a particle simulator based on the ideas behind the 'd3-force' 'JavaScript' library. 'particles' implements all forces defined in 'd3-force' as well as others such as vector fields, traps, and attractors. |
| Authors: | Thomas Lin Pedersen [aut, cre], Andrei Kashcha [ctb] |
| Maintainer: | Thomas Lin Pedersen <[email protected]> |
| License: | MIT + file LICENSE |
| Version: | 0.2.3.9000 |
| Built: | 2024-11-06 03:42:30 UTC |
| Source: | https://github.com/thomasp85/particles |
This force repositions the particles at each generation so they are centered around (0,0). It does not affect the velocity of the particles and are thus mainly a guard against the whole body of particles drifting off.
The following parameters defines the training of the force and can be passed
along a call to wield()
x : The x position to center around (tidy eval)
y : The y position to center around (tidy eval)
Other forces:
collision_force,
field_force,
link_force,
manybody_force,
map_force,
mean_force,
random_force,
reset_force,
trap_force,
x_force,
y_force
This force pushes overlapping particles apart by assigning a radius to each particle, treating them as circles, and searches for overlaps through an optimised quad tree algorithm.
The following parameters defines the training of the force and can be passed
along a call to wield()
strength : A dampening of the repulsion between overlapping circles. This
allows the force to iterate towards the optimal solution through iterative
relaxation. Should be a number between 0 and 1. Defaults to 0.7
radius : The radius of each particle. Defaults to 1 (tidy eval)
n_iter : The number of iterations to perform in the iterative relaxation.
Defaults to 1.
Other forces:
center_force,
field_force,
link_force,
manybody_force,
map_force,
mean_force,
random_force,
reset_force,
trap_force,
x_force,
y_force
This constraint requires children to be positioned at a certain side of their parent and with a certain distance. It can be used to enforce a layering of particles for e.g. DAG and tree layouts.
The following parameters defines the training of the constraint and can be
passed along a call to impose()
distance : The minimum orthogonal distance to the parent. Default to 0,
meaning that children are only required to be positioned to the specific
side of their parent. (tidy eval)
angle : The direction the children should be enforced to be relative to
their parent. Defaults to -pi/2 which is equivalent to down. (tidy eval)
Other constraints:
infinity_constraint,
path_constraint,
polygon_constraint,
velocity_constraint,
x_constraint,
y_constraint
This is the function that move the simulation forward in time. It is possible
to either specify the number of steps that should be simulated or let the
simulation terminate as alpha_min is reached. Note that some values of
alpha and alpha_target does not allow alpha to converge to alpha_min so
letting the simulation self-terminate can result in an infinite loop. The
default settings will result in alpha_min being reached in 300 generations.
evolve(simulation, steps = NULL, on_generation = NULL, ...)evolve(simulation, steps = NULL, on_generation = NULL, ...)
simulation |
A simulation object |
steps |
The number of generations to progress or a function getting the
simulation object and returns |
on_generation |
A function to be called after each generation has been progressed. The function will get the current state of the simulation as the first argument. If the function returns a simulation object it will replace the current simulation from the next generation. In the case of any other return type the return will be discarded and the function will have no effect outside its side-effects. |
... |
Additional arguments to |
Each generation in the simulation progress in the following manner:
Check whether the specified number of generations has been reached
Check whether alpha_min has been reached
If either 1. or 2. is true, terminate the simulation
Apply the forces on the current particle positions and velocities in the order they have been added
Reduce the velocity according to the given velocity_decay
Update the position and velocity based on any provided constraints
Calculate the new particle positions based on the new velocity
If given, call the on_generation function.
A simulation object with updated positions and velocities
graph <- tidygraph::create_notable('folkman') sim <- graph |> simulate() |> wield(link_force) |> wield(manybody_force) # Take 5 steps and tell about it sim |> evolve(5, function(sim) { cat('Generation: ', evolutions(sim), '\n', sep = '') }) # Run evolution until alpha_min is reached sim |> evolve(NULL)graph <- tidygraph::create_notable('folkman') sim <- graph |> simulate() |> wield(link_force) |> wield(manybody_force) # Take 5 steps and tell about it sim |> evolve(5, function(sim) { cat('Generation: ', evolutions(sim), '\n', sep = '') }) # Run evolution until alpha_min is reached sim |> evolve(NULL)
This force adjusts the velocity of particles based on a supplied vector field. The vector field can either be specified using x and y velocities, or angle and magnitude. Velocity adjustments are calculated based on a bilinear interpolation.
The following parameters defines the training of the force and can be passed
along a call to wield()
x : A matrix giving the velocity in the x direction at each grid point
y : A matrix giving the velocity in the y direction at each grid point
angle : A matrix giving the direction of the velocity at each grid point.
Will only be considered if x and y are missing.
vel : A single numeric or a matrix of the same dimensions as angle
giving the magnitude of velocity at each grid point.
xlim : The coordinate span of the vector field in the x direction.
ylim : The coordinate span of the vector field in the y direction.
Other forces:
center_force,
collision_force,
link_force,
manybody_force,
map_force,
mean_force,
random_force,
reset_force,
trap_force,
x_force,
y_force
These functions are passed to the simulation and defines how the position and velocity of the particles are initiated. The default is to lay out the nodes in a phyllotactic arrangement (think sunflower seeds) and with no velocity, which is also the default in d3-force.
phyllotactic_genesis(radius = 10, angle = pi * (3 - sqrt(5))) predefined_genesis(x, y, x_vel = 0, y_vel = 0) bigbang_genesis(vel_min = 0, vel_max = 1) aquarium_genesis(width = 10, height = 10, vel_min = 0, vel_max = 1) petridish_genesis(max_radius = 10, vel_min = 0, vel_max = 1)phyllotactic_genesis(radius = 10, angle = pi * (3 - sqrt(5))) predefined_genesis(x, y, x_vel = 0, y_vel = 0) bigbang_genesis(vel_min = 0, vel_max = 1) aquarium_genesis(width = 10, height = 10, vel_min = 0, vel_max = 1) petridish_genesis(max_radius = 10, vel_min = 0, vel_max = 1)
radius |
The radius modifier (will be multiplied by the square root of the index of the particle) |
angle |
The angular difference between two adjacent particles |
x, y
|
The columns holding (or value of) the position coordinates |
x_vel, y_vel
|
The columns holding (or value of) the velocity verlets |
vel_min, vel_max
|
The bounds of the uniformly distributed velocities |
width, height
|
The size of the rectangle holding the particles |
max_radius |
The size of the disc. |
A function that takes the particle graph and returns a list with a position and velocity element, each holding a matrix with two columns and a row for each particle giving the x and y position and velocity respectively.
phyllotactic_genesis(): Initiates particles in a phyllotactic arrangement with zero velocity
predefined_genesis(): Uses information from the node data to set position and velocity.
bigbang_genesis(): Initiates particles at center position and a random velocity
aquarium_genesis(): Places particles randomly in a rectangle and gives them a random velocity
petridish_genesis(): Places particles randomly on a disc and gives them a random velocity
# A contrieved example graph <- tidygraph::create_notable('bull') genesis <- phyllotactic_genesis() genesis(graph) # Usually used as an argument to simulate graph |> simulate(setup = phyllotactic_genesis())# A contrieved example graph <- tidygraph::create_notable('bull') genesis <- phyllotactic_genesis() genesis(graph) # Usually used as an argument to simulate graph |> simulate(setup = phyllotactic_genesis())
This function adds a new force/constraint to the simulation and trains the it
on the current particle graph. The parameters passed on to the training are
using tidy evaluation from the rlang package. Depending on the
force/constraint the data getting referenced is either the node or the edge
data of the particle graph. Both forces and constraints manipulate position
and velocity of the particles but they differ in when the are applied during
a generation. First forces are applied sequentially and the resulting
velocity is added to the resulting position after velocity_decay has been
applied. After this operation any constraint is imposed on the results. In
general, forces tends to calculate velocity adjustments, while constraints
modify position and velocity directly, but this difference is not in any way
enforced.
impose(simulation, constraint, ..., name, include = TRUE) reimpose(simulation, name, ...) unimpose(simulation, name) wield(simulation, force, ..., name, include = TRUE) rewield(simulation, name, ...) unwield(simulation, name)impose(simulation, constraint, ..., name, include = TRUE) reimpose(simulation, name, ...) unimpose(simulation, name) wield(simulation, force, ..., name, include = TRUE) rewield(simulation, name, ...) unwield(simulation, name)
simulation |
A simulation object |
constraint |
A constraint object |
... |
Parameters passed on to the training of the force or constraint |
name |
The name of the force. For use when accessing the force at a later stage. If no name is given the force is accessible by its index in the stack. |
include |
The particles to be affected by this force. Defaults to every particle in the simulation (tidy eval) |
force |
A force object |
wield() and impose() adds forces and constraints to the simulation
respectively. unwield() and unimpose() removes forces and constraints
based on name or index. rewield() and reimpose() modifies existing forces
and constraints based on name or index and retrains them.
A simulation with the force or constraint added
graph <- tidygraph::create_notable('folkman') graph |> simulate() |> wield(link_force)graph <- tidygraph::create_notable('folkman') graph |> simulate() |> wield(link_force)
This constraint keeps particles inside of a defined area by positioning exiting particles on the other side of the area. In effect this makes particles that moves outside the upper bound reenter at the lower bound and vice versa.
The following parameters defines the training of the constraint and can be
passed along a call to impose()
xlim : The left and right bound of the area
ylim : The upper and lower bound of the area
Other constraints:
dominator_constraint,
path_constraint,
polygon_constraint,
velocity_constraint,
x_constraint,
y_constraint
This force works between linked particles and either attracts or repel them from each other depending on the value of the strength and distance parameters. The force is stronger the longer the linked particles are from each other, mimicking the mechanics of a rubber band.
The following parameters defines the training of the force and can be passed
along a call to wield()
strength : The attractive force between the linked particles. The default
weighs edges from low-degree particles higher
(strength = 1 / (min(degree(from), degree(to)))). (tidy eval)
distance : The desired distance between linked particles. Defaults to
30 (tidy eval)
n_iter : The number of iteration towards the optimal solution per
generation. Higher values leads to faster convergence (measured in number
of generations) at the expense of longer computation time per generation.
Defaults to 1.
Other forces:
center_force,
collision_force,
field_force,
manybody_force,
map_force,
mean_force,
random_force,
reset_force,
trap_force,
x_force,
y_force
This force implements a n-body simulation using the Barnes-Hut approximation for improved performance. An n-body simulation calculates attraction or repulsion between all particles in a system based on their relative distances and each particles capacity and can thus mimick gravity or electrostatic repulsion.
The following parameters defines the training of the force and can be passed
along a call to wield()
strength : The attractive or repulsive force of the particles. If
positive the particle attracts, if negative the particle repulses. The
default is -30. (tidy eval)
theta : The Barnes-Hut criterion governing the precision of the
approximation. If 0, no approximation is made. Defaults to 0.9.
min_dist : A lower distance threshold below which the forces will be
damped, in order to avoid explosive forces when two particles gets very
near each other.
max_dist : A distance threshold above which the forces between particles
are ignored. Using this will result in more local changes.
Other forces:
center_force,
collision_force,
field_force,
link_force,
map_force,
mean_force,
random_force,
reset_force,
trap_force,
x_force,
y_force
In mathematics, maps are a functions that translates its input into new values. In the context of particles a map is a translation function that translates the current particle positions to a new one
Normally a map has no notion of velocity — it simply translates positions. In
particles it is possible to decide whether positions should be modified
directly or whether the translation magnitude should be added to the velocity
verlet using the fixed parameter.
The following parameters defines the training of the force and can be passed
along a call to wield()
map : A function that accepts the particle position matrix and returns
the new positions in the same format.
fixed : Logical. Should position be modified directly (TRUE) or should
the translation be added to the velocity verlet (FALSE)
Other forces:
center_force,
collision_force,
field_force,
link_force,
manybody_force,
mean_force,
random_force,
reset_force,
trap_force,
x_force,
y_force
This force takes the mean of all the neighbors (in the graph sense) of a particle (and optionally itself) and applies it to itself.
The following parameters defines the training of the force and can be passed
along a call to wield()
include_self : Should the velocity of itself be included in the mean
calculation
mode : How should neighbors be found? 'all' uses all edges. 'out'
only uses outbound edges, and 'in' only uses inbound edges. Ignored for
undirected particle graphs
Other forces:
center_force,
collision_force,
field_force,
link_force,
manybody_force,
map_force,
random_force,
reset_force,
trap_force,
x_force,
y_force
This constraint repositions particles to their closest point along a given path and sets their velocity to zero.
The following parameters defines the training of the constraint and can be
passed along a call to impose()
path : A two column matrix giving the path, or a
list of matrices to use multiple disconnected paths.
closed : Should the path close on itself. Defaults to FALSE
Other constraints:
dominator_constraint,
infinity_constraint,
polygon_constraint,
velocity_constraint,
x_constraint,
y_constraint
This constraint prevents particles from moving outside of one or more polygons. If a particle ventures outside it will be moved back to its closest point inside the specified polygon(s) and have its velocity set to zero.
The following parameters defines the training of the constraint and can be
passed along a call to impose()
polygon : A two column matrix giving the polygon, or a list of matrices
to use multiple polygons. Overlapping polygons will be subtracted from each
other so it is possible to define polygons with holes.
Other constraints:
dominator_constraint,
infinity_constraint,
path_constraint,
velocity_constraint,
x_constraint,
y_constraint
This force applies a random velocity modification to all particles. The modification is uniformly distributed and bound be the parameters provided during initialisation.
The following parameters defines the training of the force and can be passed
along a call to wield()
xmin, xmax : The bounds of the modification in the horizontal direction
ymin, ymax : The bounds of the modification in the vertical direction
Other forces:
center_force,
collision_force,
field_force,
link_force,
manybody_force,
map_force,
mean_force,
reset_force,
trap_force,
x_force,
y_force
This force resets the velocity of particles at each generation. It can be
used if each generation should start from the same foundation rather than
accumulate as the simulation evolve. Particles where the parameters evaluates
to NA will ignore this force.
The following parameters defines the training of the force and can be passed
along a call to wield()
xvel : The x-velocity to reset to at each generation (tidy eval)
yvel : The y-velocity to reset to at each generation (tidy eval)
Other forces:
center_force,
collision_force,
field_force,
link_force,
manybody_force,
map_force,
mean_force,
random_force,
trap_force,
x_force,
y_force
This function initiates a simulation based on the provided graph and
parameters. Any graph structure with a tidygraph::as_tbl_graph() method
is supported as input. This function does not start the simulation but merely
sets it up.
simulate( graph, alpha = 1, alpha_min = 0.001, alpha_decay = 1 - alpha_min^(1/300), alpha_target = 0, velocity_decay = 0.4, setup = phyllotactic_genesis(), ... ) is.simulation(x) record(simulation, ...) clear_history(simulation) get_history(simulation, age = -1) history_length(simulation) reheat(simulation, alpha) particles(simulation) position(simulation) velocity(simulation) evolutions(simulation)simulate( graph, alpha = 1, alpha_min = 0.001, alpha_decay = 1 - alpha_min^(1/300), alpha_target = 0, velocity_decay = 0.4, setup = phyllotactic_genesis(), ... ) is.simulation(x) record(simulation, ...) clear_history(simulation) get_history(simulation, age = -1) history_length(simulation) reheat(simulation, alpha) particles(simulation) position(simulation) velocity(simulation) evolutions(simulation)
graph |
A graph in a format supported by tidygraph |
alpha |
The starting alpha value. See Details. |
alpha_min |
The minimum alpha value after which the simulation is terminated. See Details. |
alpha_decay |
The speed at which the alpha value decreases. See Details. |
alpha_target |
The alpha value that alpha drifts towards. See Details. |
velocity_decay |
The dampening factor of the system. See Details. |
setup |
A function that takes the particle graph and returns a start
position and velocity to each particle. |
... |
Additional parameters for the simulation (currently ignored) |
x, simulation
|
A simulation object |
age |
The version to retrieve. Positive numbers count from the
beginning, while negative numbers counts backwards from current version.
Defaults to |
A simulation in the context of the particles package is a series of
equidistant steps where the velocity and position of each particle is
updated. A few global rules applies to this cycle irrespectively of the
forces added to the simulation. Once a simulation is initiated an alpha
value is defined (defaults to 1). At each step this alpha value is
decreased according to its distance to the alpha_target (defaults to 0)
and alpha_decay (defaults to ~0.023). Once the alpha value gets below
alpha_min (defaults to 0.001) the simulation seizes to take additional
steps. The default values is adapted from the d3-force implementation and
corresponds to 300 steps. Conceptually the alpha progression can be seen
as a cooling off of the system as the value decreases quickly in the
beginning and then slowly reach the target value. If it is not intended to
have a system that cools off, simply set the alpha_target value to the same
as alpha. At each step, after the new particle velocities has been
calculated but before they have been applied to the positions, a dampening
factor (velocity_decay) is applied in order to simulate the gradual loss
of momentum. If this is not intended for the simulation, simply set the value
to 0.
A simulation object
record(): Save the current state in the simulation's history
clear_history(): Clear the current history from the simulation
get_history(): Retrieve a simulation from the history
history_length(): Get the number of versions stored in the history of the simulation
reheat(): set the cooling of the simulation to a new value
particles(): Extract the particle graph from a simulation
position(): Extract the position coordinates from a simulation
velocity(): Extract the velocity verlets from a simulation
evolutions(): Get the number of generations the simulation has undergone
graph <- tidygraph::create_notable('folkman') graph |> simulate()graph <- tidygraph::create_notable('folkman') graph |> simulate()
The particles that are modelled in a simulation are encoded as a tbl_graph,
giving support for the particles as well as their interactions (nodes and
edges in graph parlor). A simulation supports a subset of the tidygraph/dplyr
verbs in order to allow modification of the particles after they have been
included in the simulation. In general it is possible to add and remove
particles and interactions as well as modify the metadata associated with
them. The API follows the tidygraph API where activate() is used to select
either particles or interactions and subsequent operations are thus related
to the last activated datatype. The simulation is automatically retrained
after modifying the state of the particles and their interactions.
add_particles(.data, ..., interactions = NULL, setup = NULL) replace_particles(.data, particles, setup = NULL) add_interaction(.data, ...)add_particles(.data, ..., interactions = NULL, setup = NULL) replace_particles(.data, particles, setup = NULL) add_interaction(.data, ...)
.data |
A simulation object |
... |
Parameters passed on to the main verbs in tidygraph/dplyr |
interactions |
A data.frame of interactions/edges to add along with the particles |
setup |
A function to calculate the starting conditions for the
particles. It receives all particles with the current position and
velocity encoded in the |
particles |
A |
A simulation object
dplyr::mutate(), dplyr::mutate_at(), dplyr::mutate_all(),
dplyr::filter(), dplyr::slice(), tidygraph::activate(),
tidygraph::bind_nodes(), tidygraph::bind_edges()
This force creates a trap based on any type of polygon that attracts particles as long as they are outside the polygon, while leaving particles inside the polygon unaffected. The trap as such has no walls and particles are allowed to leave it, but they will be pulled back as soon as they exits the polygon.
The following parameters defines the training of the force and can be passed
along a call to wield()
polygon : A two column matrix giving the corners of the polygon, or a
list of matrices to use multiple polygons. If multiple polygons are
overlapping it is considered a hole.
strength : The attractive force applied to the particle. Particles are
attracted towards the closest part of the polygon, rather than the center,
and the attraction is stronger for particles moving away from the polygon
than for those moving towards it. (tidy eval)
min_dist : A lower distance threshold below which the strength is not
increased. The attraction of the trap falls of with the square of the
distance to the particle, so particles close by can get an enormous
attraction unless this threshold is set (so much that the shoot out of the
other side of the trap).
distance_falloff : How should the attractive force deteriorate with the
distance between the polygon and the particle. Defaults to 2 (quadratic
falloff) (tidy eval)
Other forces:
center_force,
collision_force,
field_force,
link_force,
manybody_force,
map_force,
mean_force,
random_force,
reset_force,
x_force,
y_force
This constraint puts bounds on the magnitude of velocity a particle can have. Particles where either end of the bound is NA ignores the constraint. If a particle with no velocity is forced to have a velocity the direction will be random.
The following parameters defines the training of the constraint and can be
passed along a call to impose()
v : The velocity allowed for the particle. (tidy eval)
vmin : The lowest permissible velocity. If NULL then v will be used. (tidy eval)
vmax : The highest permissible velocity. If NULL then v will be used. (tidy eval)
Other constraints:
dominator_constraint,
infinity_constraint,
path_constraint,
polygon_constraint,
x_constraint,
y_constraint
This constraint simply prevents particles from moving in the x direction. For
particles where the constraint evaluates to NA this constraint is ignored.
If the constraint is enforced the velocity in the x direction will be set to
0.
The following parameters defines the training of the constraint and can be
passed along a call to impose()
x : The position on the x-axis to fix to. (tidy eval)
xmin : The lowest permissible x-value. If NULL then x will be used. (tidy eval)
xmax : The highest permissible x-value. If NULL then x will be used. (tidy eval)
Other constraints:
dominator_constraint,
infinity_constraint,
path_constraint,
polygon_constraint,
velocity_constraint,
y_constraint
This force simply pulls particles towards a fixed position on the x-axis.
The following parameters defines the training of the force and can be passed
along a call to wield()
strength : The strength with which the attraction occurs (tidy eval)
x : The position on the x-axis to pull towards. (tidy eval)
Other forces:
center_force,
collision_force,
field_force,
link_force,
manybody_force,
map_force,
mean_force,
random_force,
reset_force,
trap_force,
y_force
This constraint simply prevents particles from moving in the y direction. For
particles where the constraint evaluates to NA this constraint is ignored.
If the constraint is enforced the velocity in the y direction will be set to
0.
The following parameters defines the training of the constraint and can be
passed along a call to impose()
y : The position on the y-axis to fix to. (tidy eval)
ymin : The lowest permissible y-value. If NULL then y will be used. (tidy eval)
ymax : The highest permissible y-value. If NULL then y will be used. (tidy eval)
Other constraints:
dominator_constraint,
infinity_constraint,
path_constraint,
polygon_constraint,
velocity_constraint,
x_constraint
This force simply pulls particles towards a fixed position on the y-axis.
The following parameters defines the training of the force and can be passed
along a call to wield()
strength : The strength with which the attraction occurs (tidy eval)
y : The position on the y-axis to pull towards. (tidy eval)
Other forces:
center_force,
collision_force,
field_force,
link_force,
manybody_force,
map_force,
mean_force,
random_force,
reset_force,
trap_force,
x_force