Moving agents

[hsp]

I am new to Mesa. Not new to Python. Nor to ABM.
Looking at the examples it appears that most of the models are based on cellular-automata style modelling, where agents are not moving, like flocks of birds or pedestrians. Is that observation correct? If not, please, can you provide references to such models?
Furthermore, are examples available where agents are moving along infrastructure networks - for example by application of shortest path's.
I do apologize for asking such a basic and open question.
/Hans

[aiwhoo]

One recent example that's being actively integrated into mesa is shown here: #1147

You're right that most examples don't utilize the continuous space. The examples are lacking. The best example is boid flocker (https://github.com/projectmesa/mesa/tree/main/examples/boid_flockers); however, I feel the Continuous Space has been neglected since the style is completely different than the other examples.

[rht]

OP didn't mention the requirement for modelling a continuous space. Movement is a bread and butter in a lot of the examples.
E.g.

• Epstein civil violence

  mesa/examples/epstein_civil_violence/epstein_civil_violence/agent.py

  Lines 91 to 93 in c30ab5e

  	if self.model.movement and self.empty_neighbors:
  	new_pos = self.random.choice(self.empty_neighbors)
  	self.model.grid.move_agent(self, new_pos)

• Schelling

  mesa/examples/schelling/model.py

  Line 33 in c30ab5e

  self.model.grid.move_to_empty(self)

• and so on

[aiwhoo]

Good point! I'm used to the continuous space so I incorrectly made an assumption.

I think it holds though that we need more examples in the examples directory! =D

[rht]

The 2 examples I mentioned already served their purpose for showcasing movements.

[aiwhoo]

At the core, yes, this is understood by someone that really knows programming and modeling well; however, there’s ambiguity for beginners. Particularly in scenarios with defined path structures or following behaviors or desired destinations. With Mesa’s counterpart netlogo, there’s hundreds of coding examples that get beginners into it easily; while mesa only has 16 examples, which may not resonate and left the beginners frustrated and returning back to other modeling platforms.

[tpike3]

@aiwhoo always happy to have more examples to merge.

On that note we are working with Santa Fe to set up a tutorial class through complexity explorer on Mesa. The plan is for second portion to walk through how to contribute an example (thanks @rht).

@rht based on our other conversation my thought is to do sugar and space traders from sugarscape as the example...what do you think?

[rht]

@rht based on our other conversation my thought is to do sugar and space traders from sugarscape as the example...what do you think?

Sounds good. You mentioned that those models depend on #1142 being merged. But #1142 has several design issues that have yet to be resolved. But there is always a workaround by just adding the by-type activation from scratch in the model's model.py.

[hsp]

Eventually a full integration of Geo-data and GIS/Spatial modelling capabilities would be great for creation of real-world models. As you probably all know there is a lot of very evolved options available. GDAL, OGR, Shapely etc. Inclusions of features (points, lines, and polygons), networks and raster’s would be good. Do you happen to know any examples?

[jackiekazil]

It already exists -- https://github.com/Corvince/mesa-geo