Mountain Runners

Let's consider a race to the top of a mountain. Runners have been scattered about a map in some low lying areas and need to find the best path up to the peak.

We'll use Pathfinding.AStar and a Pathfinding.PenaltyMap to simulate this.

Setup

using Agents, Agents.Pathfinding
using Random
using FileIO # To load images you also need ImageMagick available to your project

@agent struct Runner(GridAgent{2}) end

Our agent, as you can see, is very simple. Just an id and position provided by @agent. The rest of the dynamics of this example will be provided by the model.

function initialize(map_url; goal = (128, 409), seed = 88)
    # Load an image file and convert it do a simple representation of height
    heightmap = floor.(Int, convert.(Float64, load(download(map_url))) * 255)
    # The space of the model can be obtained directly from the image.
    # Our example file is (400, 500).
    space = GridSpace(size(heightmap); periodic = false)
    # The pathfinder. We use the `MaxDistance` metric since we want the runners
    # to look for the easiest path to run, not just the most direct.
    pathfinder = AStar(space; cost_metric = PenaltyMap(heightmap, MaxDistance{2}()))
    model = StandardABM(
        Runner,
        space;
        agent_step!,
        rng = MersenneTwister(seed),
        properties = Dict(:goal => goal, :pathfinder => pathfinder)
    )
    for _ in 1:10
        # Place runners in the low-lying space in the map.
        runner = add_agent!((rand(abmrng(model), 100:350), rand(abmrng(model), 50:200)), model)
        # Everyone wants to get to the same place.
        plan_route!(runner, goal, model.pathfinder)
    end
    return model
end
initialize (generic function with 1 method)

The example heightmap we use here is a small region of countryside in Sweden, obtained with the Tangram heightmapper.

Dynamics

With the pathfinder in place, and all our runners having a goal position set, stepping is now trivial.

agent_step!(agent, model) = move_along_route!(agent, model, model.pathfinder)
agent_step! (generic function with 1 method)

Let's Race

Plotting is simple enough. We just need to use the InteractiveDynamics.abmplot for our runners, and display the heightmap for our reference. A better interface to do this is currently a work in progress.

using CairoMakie

We load the sample heightmap

map_url =
    "https://raw.githubusercontent.com/JuliaDynamics/" *
    "JuliaDynamics/master/videos/agents/runners_heightmap.jpg"
model = initialize(map_url)
StandardABM with 10 agents of type Runner
 agents container: Dict
 space: GridSpace with size (400, 500), metric=chebyshev, periodic=false
 scheduler: fastest
 properties: pathfinder, goal

and plot

const ABMPlot = Agents.get_ABMPlot_type()
function Agents.static_preplot!(ax::Axis3, p::ABMPlot)
    return scatter!(ax, model.goal; color = (:red, 50), marker = 'x')
end

abmvideo(
    "runners.mp4",
    model;
    figurekwargs = (size = (700, 700),),
    frames = 200,
    framerate = 45,
    agent_color = :black,
    agent_size = 8,
    agentsplotkwargs = (strokecolor = :white, strokewidth = 2),
    heatarray = model -> penaltymap(model.pathfinder),
    heatkwargs = (colormap = :terrain,),
)