# Maze Solver

Consider a scenario where a walker agent is stuck in a maze. Finding the shortest path through an arbitrary maze or map is simulated using a Pathfinding.AStar and its walkmap map property.

## Setup

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

The Walker agent needs no special property, just the id and position from @agent.

@agent Walker GridAgent{2} begin end

The maze is stored as a simple .bmp image, where each pixel corresponds to a position on the grid. White pixels correspond to walkable regions of the maze.

function initalize_model(map_url)
# Load the maze from the image file. White values can be identified by a
# non-zero red component
# The size of the space is the size of the maze
space = GridSpace(size(maze); periodic = false)
# Create a pathfinder using the AStar algorithm by providing the space and specifying
# the walkmap parameter for the pathfinder.
# Since we are interested in the most direct path to the end, the default
# DirectDistance is appropriate.
# diagonal_movement is set to false to prevent cutting corners by going along
# diagonals.
pathfinder = AStar(space; walkmap=maze, diagonal_movement=false)
model = ABM(Walker, space)
# Place a walker at the start of the maze
walker = Walker(1, (1, 4))
# The walker's movement target is the end of the maze.
plan_route!(walker, (41, 32), pathfinder)

return model, pathfinder
end

# Our sample walkmap
map_url =
model, pathfinder = initalize_model(map_url)
(AgentBasedModel with 1 agents of type Walker
space: GridSpace with size (41, 41), metric=chebyshev, periodic=false
scheduler: fastest, A* in 2 dimensions, orthogonal, ϵ=0.0, metric=DirectDistance)

# Dynamics

Stepping the agent is a trivial matter of calling move_along_route! to move it along it's path to the target.

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

## Visualization

Visualizing the Walker move through the maze is handled through InteractiveDynamics.abmplot.

using InteractiveDynamics
using CairoMakie

The heatarray keyword argument allows plotting the maze as a heatmap behind the agent.

abmvideo(
"maze.mp4",
model,
agent_step!;
figurekwargs = (resolution=(700,700),),
frames=60,
framerate=30,
ac=:red,
as=11,
heatarray = _ -> pathfinder.walkmap,
)