Voronoi Tessellation based Discrete Space #2084

vitorfrois · 2024-03-21T00:41:22Z

First version of Voronoi Tesselation based Discrete Space,

described on issue #1895 . This feature allows the user to build a discrete space based on a random sample of points, where neighbors are defined by Delaunay Triangulation.

More specifically, Delaunay Triangulation is a dual-graph representation of the Voronoi Tesselation. Using this algorithm, we can easily find nearest neighbors without delimiting cells edges.

The library chosen for the triangulation algorithm implementation was PyHull, a wrapper of qhull C library, which deals with spatial calculations. The choice was made considering performance, size and usability (SciPy has a similar module but much heavier).

Based on my discussion with @EwoutH on the issue thread, i thought it would be useful to inherit DiscreteSpace class.

class VoronoiGrid(DiscreteSpace):

Example

from mesa import Model
from mesa.experimental.cell_space import VoronoiGrid, CellAgent

class MyAgent(CellAgent):
    def __init__(self, unique_id, model):
        super().__init__(unique_id, model)

Users can input cells centroids

points = [
    [0, 0], [0, 1], [0, 2], 
    [1, 0], [1, 1], [1, 2],
    [2, 0], [2, 1], [2, 2]]
grid = VoronoiGrid(centroids_coordinates=points, capacity=1, random=model.random)

and cells neighborhoods are defined by Delaunay triangulation

agent = MyAgent(1, model)
cell = grid.select_random_empty_cell()
agent.move_to(cell)
print(cell)
print(cell.neighborhood())

Cell([1, 0], [<__main__.MyAgent object at 0x7f2177268450>])
CellCollection({Cell([0, 1], []): [], Cell([1, 1], []): [], Cell([0, 0], []): [], Cell([2, 1], []): [], Cell([2, 0], []): []})

Next Steps

Definitely, next steps involve computing the boundaries of each cell using Voronoi Tesselation. This allow us to define capacity of each cell based on its area/volume.

EwoutH · 2024-03-23T11:21:10Z

Thanks, it's a great start! Using DelaunayTri is a smart approach.

I guess move_to currently move an agent to the cells centroid? While that is logical behavior, I would also like to see an method in which agents are moved to A) the closest point in the cell from their current position and B) a random spot in the cell. This could be a separate PR.

Another thing that it's needed in the long term is a method to check in which cell a point (and thus an agent) is.

@wang-boyu I would also appreciate a review from your mesa-geo expertise.

For the currently functionality, I would like some additional docstring and unittests. Then you can decide if you want to add further functionality to this PR, or merge this first and follow up with new PRs.

Edit: One thought, also for the other maintainers: Would we need separate Discrete and Continuous Voronoi spaces?

EwoutH · 2024-07-11T12:06:23Z

mesa/experimental/cell_space/voronoi.py

+class VoronoiGrid(DiscreteSpace):
+    def __init__(
+        self,
+        centroids_coordinates: Sequence[Sequence[float]],


Would this work with Shapely Points? That might be very useful for Mesa geo applications (cc @wang-boyu).

Just add with adding a simple test case for this, maybe it will just work.

mesa/experimental/cell_space/voronoi.py

wang-boyu · 2024-07-12T02:31:24Z

Thanks for the PR. I have to say that I'm still a bit confused about why this is based on discrete space, not continuous space (as I mentioned in #1895 (comment) and #1895 (comment)).

As a concrete example, consider this simple space:

Cell 1	Cell 2	Cell 3
O	?	X

If it is a Voronoi space, should Cell 2 be split into halves, where left half belongs to agent of type O and right half belongs to agent of type X? How does this work in discrete space?

tpike3 · 2024-07-12T12:17:03Z

Apologies, a little late to this discussion, this is really cool @vitorfrois.

As dependencies and keeping Mesa lightweight is a constant struggle, but form a very superficial look is there a reason we wont use networkx (https://networkx.org/documentation/stable/reference/algorithms/voronoi.html) and this has a corresponding example that may work for mesa_geo (https://networkx.org/documentation/stable/auto_examples/geospatial/plot_delaunay.html)?

Let me know what you think/ what I am missing?

vitorfrois · 2024-07-12T12:37:14Z

@wang-boyu

I have to say that I'm still a bit confused about why this is based on discrete space, not continuous space. How does this work in discrete space?

I'm considering the discrete space for Voronoi Tesselation as a static/non changing graph, where neighbors are obtained by Delaunay Triangulation. As @EwoutH said, it is only a special case for HexGrid where cell are not equally spaced.

I think its more about a choice of implementation where I considered mainly the Cholera example. As I suggested yesterday,

since the result of Delaunay Triangulation is a graph, we can use NetworkGrid to represent it, and in the continuous case, implement in the Mesa Geo using Shapely.

Does that make sense?

vitorfrois · 2024-07-12T12:43:55Z

@tpike3 thanks for the comment. That's more or less what I'm thinking. In Mesa Geo, we have opportunity to do it better with Shapely

wang-boyu · 2024-07-13T01:09:19Z

I'm considering the discrete space for Voronoi Tesselation as a static/non changing graph, where neighbors are obtained by Delaunay Triangulation.

Thanks for clarifying. Appreciate it.

Question: Is the Delaunay Triangulation done in continuous space?

vitorfrois · 2024-07-13T15:07:32Z

Question: Is the Delaunay Triangulation done in continuous space?

Yes, it is

While it might be desired in a specific model to have the same agent be placed in multiple spots simultaneously, the typical use case is that one agent has one position at every given moment. This commit decorates the place_agent() method in a way that it emits a warning when called with an agent which already has a location. Fixes: projectmesa#1522

uv is a fast drop-in pip, pip-compile, virtualenv etc replacement created by the creator of Ruff.

* Update Ruff to 0.3.4; apply ruff format . * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…ojectmesa#2087) This fixes - some minor grammatical errors, - an incorrect header indentation level,

updates: - [github.com/astral-sh/ruff-pre-commit: v0.3.4 → v0.3.5](astral-sh/ruff-pre-commit@v0.3.4...v0.3.5) - [github.com/asottile/pyupgrade: v3.15.0 → v3.15.2](asottile/pyupgrade@v3.15.0...v3.15.2)

* Jupyter Viz: Don't avoid interactive backend We were avoiding the interactive backend, but that isn't recommended anymore and Solara should take care of that itself. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Sets the version to 3.0.0a0 for the first Mesa 3.0 pre-release, and updates the release notes with 2.3.1 and 3.0.0a0 changelog

Don't track virtual environment files in Git

for more information, see https://pre-commit.ci

vitorfrois · 2024-07-15T18:27:52Z

Updates

Implemented Voronoi from scratch (basically copied a simple already implemented algorithm)
Now using the regions resulting from Delaunay/Voronoi Triangulation to draw Voronoi cells in JupyterViz. As the cell is an important in voronoi diagram, I added an attribute cell_coloring_property to VoronoiGrid. When drawing the grid, the polygon is drawn with plt.fill(polygon, alpha=cell.properties[space.cell_coloring_property]), so one can tweak the cell coloring property to adjust the visualization. In the example, the cell coloring property is cases ratio, which give darker shades on cells with more cases.
Altough, could not cover visualization/components/matplotlib.py with tests :(
It was simpler and more visual to directly implement Voronoi cells since they were computed anyway, instead of the visualization using NetworkX, mentioned earlier. It is not that performant though

This gives us the example above

EwoutH

This is really outstanding, Vítor. That space visualization especially, it really offers a lot of value.

Small minor comments. I would also like to see more test coverage, if you have the chance.

I’m going to approve conceptually, especially considering the cell space is still experimental. @rht are you able to do a technical code review and merge if it looks good?

CC @quaquel, you might also find this interesting (and have some comments).

pyproject.toml

mesa/experimental/cell_space/voronoi.py

vitorfrois · 2024-07-21T12:20:20Z

Omg, sorry I did not notice the license. I copied like 70% of it's implementation, changing variables, function names and other minor things.

…

On Sun, Jul 21, 2024, 08:44 rht ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In mesa/experimental/cell_space/voronoi.py <#2084 (comment)>: > @@ -0,0 +1,264 @@ +from collections.abc import Sequence +from itertools import combinations +from random import Random + +import numpy as np + +from mesa.experimental.cell_space.cell import Cell +from mesa.experimental.cell_space.discrete_space import DiscreteSpace + + +class Delaunay: + """ + Class to compute a Delaunay triangulation in 2D + ref: http://github.com/jmespadero/pyDelaunay2D How much percentage of the code is based on this GitHub repo? Their license is GPLv3, and we actually can't copy their code because there will be a licensing issue. — Reply to this email directly, view it on GitHub <#2084 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALBWUBC63DYQCEJC5TPSGOLZNONJHAVCNFSM6AAAAABFAPRYMCVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDCOJQGMYDSMZXGM> . You are receiving this because you were mentioned.Message ID: ***@***.***>

rht · 2024-07-21T13:35:40Z

There are 2 ways to proceed:

Use ChatGPT to do code laundry, i.e. ask it to do Delaunay triangulation code. It probably is going to be based on that GPLv3 repo, but you can blame ClosedAI for that instead.
Use SciPy, but have SciPy to be an optional dependency, only imported when the user needs VoronoiGrid

I'm more in favor of option 2, because option 1 is rather shady.

EwoutH · 2024-07-21T18:37:57Z

Yeah the license issue is a big one. Another (far fetched) option is to contact the maintainer to see if he’s willing to publish on another, more permissive license.

EwoutH · 2024-07-25T15:18:29Z

@vitorfrois I would like to try to contact the maintainers. Do you want to write them a message or would you like me to do it?

EwoutH · 2024-07-29T13:49:04Z

I can do it tomorrow, then hopefully we can move this forward next week!

I would like to include it in the next 3.0 alpha release.

EwoutH · 2024-08-01T14:23:09Z

I reached out Tuesday, hopefully we will get an answer soon: jmespadero/pyDelaunay2D#7

EwoutH

You know what, I think this is fine. It's just math in a specific order.

Also, this is in the experimental space. We can break or remove anything from here, and iterate fast on it.

Let's get this moving. If you remove the dependency we're good to go, and I will merge.

Edit: There are a few minor open comments. Would be nice to have those resolved.

EwoutH · 2024-08-30T14:47:09Z

Merged! Congratulations @vitorfrois your first PR is in Mesa - and it's a remarkable one.

Thanks for your effort and patience.

It will be included in the Mesa 3.0.0a3 release later today.

vitorfrois added 3 commits March 20, 2024 17:13

Implemented Voronoi based discrete space

fa8c15d

Using coordinates instead of dimensions and density

301a5ff

deleted comments

a8a17ab

EwoutH requested review from quaquel, Corvince and EwoutH March 23, 2024 11:12

EwoutH added feature Release notes label experimental Release notes label labels Mar 23, 2024

EwoutH requested a review from wang-boyu March 23, 2024 11:19

vitorfrois added 2 commits March 23, 2024 18:42

incremented docstrings

7e4d79e

added voronoi tests

8a892ee

EwoutH mentioned this pull request Mar 24, 2024

Add Voronoi tessellation space #1895

Closed

2 tasks

vitorfrois mentioned this pull request Mar 26, 2024

Use new VoronoiGrid discrete space in examples projectmesa/mesa-examples#112

Open

EwoutH reviewed Jul 11, 2024

View reviewed changes

mesa/experimental/cell_space/voronoi.py Show resolved Hide resolved

EwoutH reviewed Jul 11, 2024

View reviewed changes

mesa/experimental/cell_space/voronoi.py Outdated Show resolved Hide resolved

EwoutH mentioned this pull request Jul 11, 2024

Cholera Voronoi projectmesa/mesa-examples#118

Open

puer-robustus and others added 6 commits July 15, 2024 13:07

ci: Use uv pip for faster build (projectmesa#2038)

6b8e760

uv is a fast drop-in pip, pip-compile, virtualenv etc replacement created by the creator of Ruff.

test: Remove place_agent duplicate warnings

63bffa4

docs: Fixes typos and header level error in introductory tutorial (pr…

c72dbff

…ojectmesa#2087) This fixes - some minor grammatical errors, - an incorrect header indentation level,

[pre-commit.ci] pre-commit autoupdate

36adaff

updates: - [github.com/astral-sh/ruff-pre-commit: v0.3.4 → v0.3.5](astral-sh/ruff-pre-commit@v0.3.4...v0.3.5) - [github.com/asottile/pyupgrade: v3.15.0 → v3.15.2](asottile/pyupgrade@v3.15.0...v3.15.2)

EwoutH and others added 12 commits July 15, 2024 13:07

Remove last remnants of visualization_old (projectmesa#2169)

ec4233a

Set version to 3.0.0a0 and update release notes (projectmesa#2167)

fb4177c

Sets the version to 3.0.0a0 for the first Mesa 3.0 pre-release, and updates the release notes with 2.3.1 and 3.0.0a0 changelog

Add .venv/ to .gitignore (projectmesa#2172)

998965f

Don't track virtual environment files in Git

Add original conference paper link to docs (projectmesa#2160)

a7090fc

added voronoi visualization

909831f

var names

981473f

voronoi visualization

5b66f73

Merge branch 'main' into frois-voronoi

554dc43

[pre-commit.ci] auto fixes from pre-commit.com hooks

0939e3c

for more information, see https://pre-commit.ci

added voronoi tests

ab1d1b6

added voronoi tests

c2d7379

EwoutH approved these changes Jul 20, 2024

View reviewed changes

pyproject.toml Outdated Show resolved Hide resolved

mesa/experimental/cell_space/voronoi.py Show resolved Hide resolved

rht reviewed Jul 21, 2024

View reviewed changes

mesa/experimental/cell_space/voronoi.py Show resolved Hide resolved

EwoutH mentioned this pull request Jul 30, 2024

More permissive license jmespadero/pyDelaunay2D#7

Open

EwoutH mentioned this pull request Aug 1, 2024

here is my project. is my Urban Traffic Simulation with GIS Data projectmesa/mesa-examples#132

Closed

EwoutH approved these changes Aug 27, 2024

View reviewed changes

vitorfrois and others added 3 commits August 27, 2024 23:08

removed pyhull dependency

f1857a6

Merge remote-tracking branch 'upstream/main' into pr/2084

b307e2b

Update test_solara_viz.py

e0ded76

EwoutH merged commit c19f53e into projectmesa:main Aug 30, 2024
9 of 12 checks passed

EwoutH changed the title ~~Voronoi Tesselation based Discrete Space~~ Voronoi Tessellation based Discrete Space Aug 30, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Voronoi Tessellation based Discrete Space #2084

Voronoi Tessellation based Discrete Space #2084

vitorfrois commented Mar 21, 2024 •

edited

Loading

EwoutH commented Mar 23, 2024 •

edited

Loading

EwoutH Jul 11, 2024 •

edited

Loading

wang-boyu commented Jul 12, 2024

tpike3 commented Jul 12, 2024

vitorfrois commented Jul 12, 2024

vitorfrois commented Jul 12, 2024

wang-boyu commented Jul 13, 2024

vitorfrois commented Jul 13, 2024

vitorfrois commented Jul 15, 2024 •

edited

Loading

EwoutH left a comment

vitorfrois commented Jul 21, 2024 via email

rht commented Jul 21, 2024

EwoutH commented Jul 21, 2024

EwoutH commented Jul 25, 2024

EwoutH commented Jul 29, 2024 •

edited

Loading

EwoutH commented Aug 1, 2024 •

edited

Loading

EwoutH left a comment •

edited

Loading

EwoutH commented Aug 30, 2024

Voronoi Tessellation based Discrete Space #2084

Voronoi Tessellation based Discrete Space #2084

Conversation

vitorfrois commented Mar 21, 2024 • edited Loading

First version of Voronoi Tesselation based Discrete Space,

Example

Next Steps

EwoutH commented Mar 23, 2024 • edited Loading

EwoutH Jul 11, 2024 • edited Loading

Choose a reason for hiding this comment

wang-boyu commented Jul 12, 2024

tpike3 commented Jul 12, 2024

vitorfrois commented Jul 12, 2024

vitorfrois commented Jul 12, 2024

wang-boyu commented Jul 13, 2024

vitorfrois commented Jul 13, 2024

vitorfrois commented Jul 15, 2024 • edited Loading

EwoutH left a comment

Choose a reason for hiding this comment

vitorfrois commented Jul 21, 2024 via email

rht commented Jul 21, 2024

EwoutH commented Jul 21, 2024

EwoutH commented Jul 25, 2024

EwoutH commented Jul 29, 2024 • edited Loading

EwoutH commented Aug 1, 2024 • edited Loading

EwoutH left a comment • edited Loading

Choose a reason for hiding this comment

EwoutH commented Aug 30, 2024

vitorfrois commented Mar 21, 2024 •

edited

Loading

EwoutH commented Mar 23, 2024 •

edited

Loading

EwoutH Jul 11, 2024 •

edited

Loading

vitorfrois commented Jul 15, 2024 •

edited

Loading

EwoutH commented Jul 29, 2024 •

edited

Loading

EwoutH commented Aug 1, 2024 •

edited

Loading

EwoutH left a comment •

edited

Loading