Plotting agents in a domain
Plotting agents in a domain#
On this page, you will see how to
.plot.abmto plot snapshots and animations of agents in a domain
plot data variables onto the agent marker size, orientation, and color
add fading trails to the agents moving around the domain
adjust the plot for periodic boundary conditions
adjust the plot domain
Complete example: Agents in a domain (SimpleFlocking model)
agents: based_on: - .creator.universe - .plot.abm - .animation.ffmpeg - .animation.fps30 # Prepare data select: x: path: agent/x transform: [ .data ] allow_failure: silent fallback: !dag_tag _skip_plot y: path: agent/y transform: [ .data ] orientation: path: agent/orientation transform: [ .data ] space_extent: path: agent/x # meta data is written to every other dataset as well transform: - getattr: [!dag_prev , attrs] - getitem: [!dag_prev , space_extent] transform: # Combine data into one dataset - xr.Dataset: - x: !dag_tag x y: !dag_tag y orientation: !dag_tag orientation tag: agents # Data encoding x: x y: y frames: time orientation: orientation hue: orientation # Define what to plot to_plot: agents: marker: wedge # Pass on space extent (used for setting limits) domain: extent: !dag_result space_extent pad: 0 # Set tail lengths, style, and decay factor tail_length: 24 tail_decay: 0.12 tail_max_segment_length: 1. # to avoid problems when crossing a boundary tail_kwargs: color: black linewidth: 0.5 alpha: .6 zorder: -10 # Colorbar settings vmin: -3.14159 vmax: +3.14159 cbar_labels: +3.14159: $+π$ +1.570795: $+π/2$ 0: $0$ -1.570795: $+π/2$ -3.14159: $-π$
Agent-based models (ABMs) are a common type of model to simulate and analyze the behaviour of individual agents
moving around in space, interacting with each other, and adapting their behaviour or even learning in the process.
Utopia uses utopya’s
.plot.abm plot (implemented in utopya as
abmplot()) to visualize the behavior of such agents.
We will use the SimpleFlocking ABM to illustrate its functionality.
The implementation of this plot is in
utopya.eval, which is where you can find details about the available arguments and their behavior:
Animated agents in a domain#
Below is an example animation of flocking dynamics in a square domain with periodic boundary conditions:
The agents’ color represents their orientation angle, and the fading trails their trajectory over a number of past steps.
Such animations can be easily generated using the
.plot.abm base plot.
The plot requires a single
xarray.Dataset or an
xarray.DataArray to be passed containing all the plotting data.
Base your plot on the universe creator (
.plot.abm, as well as an animation base configuration (see the
animations page for more details on these, including on adjusting the animation resolution):
agents_in_domain: based_on: - .creator.universe - .plot.abm - .animations.ffmpeg select: agents: path/to/agent/data
Next, specify what to plot, and where to plot the variables:
agents_in_domain: # Everything from above ... to_plot: agents: marker: wedge x: x y: y frames: time orientation: orientation hue: orientation
This plots the variables onto the corresponding dimensions.
to_plot argument is required; it can also contain information on the agent markers, colors, and size scales – see below.
Naturally, this requires the data we selected to contain variables called
orientation (which can be mapped to other variables using the corresponding keyword arguments).
Arguments on the top-level of the plot config are used as defaults for all entries within
If the individual
to_plot entries have different encodings, you can also specify those arguments there:
agents_in_domain: # ... to_plot: agents: # custom parameters for this layer x: y y: x
agents_in_domain: based_on: # As before ... select: x: path: path/to/x/coordinates transform: [.data] # Transform into a xr.DataArray y: path: path/to/y/coordinates transform: [.data] # analogously for other variables ... # Combine data into one dataset transform: - xr.Dataset: - x: !dag_tag x y: !dag_tag y # other variables ... tag: agents to_plot: agents: marker: wedge
You can turn off the time stamp by setting
Plotting frames and snapshots#
You can of course choose specific time steps to use for the animation. To do this, pass the
agents_in_domain: # Everything else as above ... frames_isel: !range 
This will plot the first ten time steps.
Alternatively, you can pass a list of frames.
If you base the plot on
animation.frames and do not pass
frames_isel, all frames will be plotted as individual images.
You can also plot snapshots instead of an animation.
To do this, add the
.plot.abm.snapshot base configuration; then, select the time frames to plot with
agents_in_domain: based_on: - .creator.universe - .plot.abm - .plot.abm.snapshot # Snapshot of the final state (default) frames_isel: -1
Agent hue, orientation, size, and markers#
As we saw above, the
hue of the agents can be used to visualize data variables. You can then specify
a colormap using the ColorManager:
agents_in_domain: # Everything as above cmap: continuous: true from_values: 0.0: darkgreen 0.333: yellow 0.666: brickred 1.0: darkgreen
Colors can also be specified e.g. as hex or rgb values. You can also simply pass the name of a colormap, use a norm, and include limits: take a look at the style section for more details.
Additionally, you can plot data dimensions onto the agent
size. The size scale and marker style of the agents is determined via the
size_scale key in the
to_plot entry, which roughly corresponds to the area of the marker in relation to the whole domain:
agents_in_domain: # Everything as above ... size: some_dimension to_plot: agents: size_scale: 0.0002 marker: fish2
marker argument can be any one of
fish (a basic fish shape), and
fish2 (a more complex fish shape); see
MARKERS for more information and other available marker paths:
AgentCollection for more possibilities on how to set up the
Trails can be useful to visualize the trajectories of the agents, especially in flocking or chemotaxis models.
You can control the length of the tail with the
tail_length entry. The aesthetics of the tail are controlled via the
agents_in_domain: tail_length: 12 tail_kwargs: color: black linewidth: 0.5 alpha: .6 zorder: -10
This sets the
tail_length to 12 frames of the animation.
alpha value is applied uniformly along the length of the tail.
To get a fading effect, add the following key:
agents_in_domain: tail_decay: 0.12
This will set the alpha value of each tail segment to
1 - tail_decay times the alpha value of the previous segment, giving an exponential fade.
Periodic boundary conditions#
tail_max_segment_length parameter is useful if you plan on drawing tails of agents that move in a periodic space.
In such a case, agent positions may jump aprubtly when crossing a boundary.
Ordinarily, this would lead to the tail segment going across the whole domain.
To avoid this, set the
tail_max_segment_length parameter to half the domain size; this typically suffices to detect jumps in x- or y- position and leads to these segments not being drawn.
(To be precise, the length refers not to that of the segment but to the differences in x- and/or y-position.)
Adjusting the plot domain#
Instead of using the
PlotHelper to adjust the x- and y-extent, we recommend using the
domain/extent key; this way,
marker size will be kept constant. This key also allows setting the padding to the border of the domain, the height, and
the aspect of the domain:
agents_in_domain: domain: extent: [0, 4, 1, 10] pad: 0.5 height: 3 # height in data units aspect: 2 # can also be 'auto' if height is not given
extent entry should be a tuple of the form
(left, right, bottom, top).
Alternatively, a 2-tuple will be interpreted as
(0, right, 0, top).
Finally, you can also add a
mode key to the
domain dictionary to control automatic deduction of boundaries.
The domain mode can be
fixed mode, all available data is inspected to derive domain bounds.
follow mode, boundaries are adjusted from frame to frame.
auto mode, will use a
fixed domain if no
extent was given, otherwise
manual mode is used.