A* Pathfinding

This library is another implementation in JavaScript of the famous A* algorithm.

• Live demo
• Documentation

Features

• Based on Graph theory for more flexibility
  • Multiple graphs support, use the same nodes instances for differents graph instances
• Graph builder helper for 2D Arrays
  • Any access order supported ([x][y] or [y][x])
  • Diagonals move support, with or without corner-cutting
  • Toric maps support (maps wrapped on itself)
  • Jump Point Search (JPS) support for faster computations on uniform cost grids. (NB: Toric support isn't yet supported with JPS)
• 3 heuristics functions availables : Manhattan, Diagonals or Euclidian
  • Custom heuristic function also supported
• Variable cost support
• Multi-threading support (with Web Workers)
• Use of Binary Heaps and Connectivity to speed up calculations

Getting Started

First of all, you'll need to include the library :

    <script src="./bin/lowlight.astar.js"></script>

You may include the minified library instead :

    <script src="./bin/lowlight.astar.min.js"></script>

Then you may create alias for convenience :

    let Astar = Lowlight.Astar

Create a graph

If you're using a custom data structure, you'll need to convert it into a Graph struture. Don't worry, it's quite easy.

Assume your datas look like this :

    let a = {id:"pallet-town", gym:null}
    let b = {id:"viridian-city", gym:"Earth badge"}
    let c = {id:"cinnabar-island", gym:"Volcano badge"}

First, start by creating a new Graph and some nodes :

    let graph = new Astar.Graph()
    a = new Astar.Node(a.id, a) //Datas are synchronized :
    b = new Astar.Node(b.id, b) //b.gym = "Earth badge"
    c = new Astar.Node(c.id, c) //c.gym = "Volcano badge"

Then you'll need to override Graph.id(id) method because graphs instances only use numbers as identifiers.

    //Override graph.id method
    graph.id = function (name) {
        return ["pallet-town", "viridian-city", "cinnabar-island"].indexOf(name)
    }

    //This way, graph.id will be compatible with any id type
    graph.id("pallet-town") //Return 0

Finally link different nodes :

    graph.edge(a, b, 1, 2) //Going from a to b cost 1 whereas going from b to a cost 2
    graph.edge(a, c, 1, null) //Going from a to b cost 1 whereas it's not possible to go from b to a

Create a graph from an array structure

For conveniency, it's possible to create directly a graph from an array structure without specific order :

YX Order

    let map =  [
        [0, 1, 0],
        [0, 1, 1],
        [0, 0, 0],
    ]

XY Order

    let map =  []
    for (let x = 0; x < 3; x++) { map[x] = []
        for (let y = 0; y < 3; y++) {
            map[x][y] = Math.random() > 0.5
    }   }

Create a configuration

To create a new Configuration, just type the following code :

    let astar = new Astar.Configuration(map)

If you want to specify options, you can pass a second argument to instanciation :

    let graph = new Astar.Configuration(map, {
        order:"yx", //Access order for 2D Array
        torus:false, //Tell if 2D Array is wrapped on itself
        diagonals:true, //Tell if diagonals moves are allowed (2D Array)
        cutting:false, //Tell if diagonals moves between two blocking squares are allowed (2D Array)
        heuristic:"euclidian", //Heuristic function name
        cost(a, b) { return b == 1 ? null : 1 }, //Cost function definition
        thread:"./bin/lowlight.astar.min.js", //If script's source is specified, multi-threading will be enabled
    })

If you want to reuse same map but with different option, just create instance with multiple options objects.

    let graph = new Astar.Configuration(map, layer1, layer2, layer3)

Cost function

Cost function takes will be called with two Node instances : source node and destination node. It must returns a number or null.

While it's possible to use Infinity and negatives values, it isn't advised because it can leads to suspicious behaviours.

Most of the time, you'll only need the second argument, but you may want to use the first one to detect transitions. For example, going from plains to mountains may cost 2 whereas going from mountains to mountains may cost only 1.

    let cost = function(a, b) {
        if (b === "sea") { return null }
        if (b === "plains") { return 1 }
        if ((a === "mountains")&&(b === "mountains")) { return 1 }
        if ((a === "plains")&&(b === "mountains")) { return 2 }
        return null
    }

Heuristic function

By default, only two-dimensional heuristic functions are defined.

To add a new one, just type the following code :

    Astar.Heuristic["myHeuristic"] = function (a, b, options) {
        return Math.abs(b.x - a.x) + Math.abs(b.y - a.y) + Math.abs(b.z - a.z)
    }

The first two arguments are source node and destination node (remember that you can store anything in a node instance). The third one is optional and may be specified as heuristicOptions in astar.path method.

Computing a path

To compute a path, you'll just need to pass a start node and a goal node.

    let path = astar.path({x:0, y:0}, {x:2, y:2})

If multi-threading option is enabled, you'll be need to specify a callback function.

    astar.path({x:0, y:0}, {x:2, y:2}, {callback(path, scores) { path.map(n => console.log(n)) }})

Note that if you're using a multiple layers graph, you may specify the layer you want to use, the default one being 0.

    astar.path({x:0, y:0}, {x:2, y:2}, {layer:1})

Using Jump Search Point (JPS)

A* JPS is pruning rules to avoid node exploration expansion. This means that it's generally much faster than classic A*, however, it comes with a major drawback : movement costs must be uniform.

You may specify that you want to use jps in options when computing a path :

    astar.path({x:0, y:0}, {x:2, y:2}, {jps:true})

Nota Bene : This only works for two-dimensionals grid.

Using static graphs

If your graph is static, you can set the static option to prevent useless computations. Before computing a path, it'll do a connectivity test to check if a path exists before computing one.

    astar.path({x:0, y:0}, {x:2, y:2}, {static:true})

You can also use it on dynamics graphs, but don't forget to rebuild connectivity by calling graph.connectivity() method each time you edit it.

Project content

Rebuild project and expanding the library

You'll need to run the following command the first time to install dependencies.

npm install

Then to rebuild project, just run the following command :

npm run build

This will update /bin files with included /src files. Although package.json (which contains "source" and "output" paths) are preconfigured, you may need to reconfigure them if you're planning to expand this library.

To include a file just use the following syntax in the "source" file :

    /* #include <path/to/file.js> */

• File minification is performed with Babel minify.
• Documentation is generated with JSDoc 3.

Although package.json (which contains "jsdoc_source", "jsdoc_output", "jsdoc_config" and "jsdoc_readme") and docs/categories.json are preconfigured, you may need to reconfigure them if you're planning to expand this library.

License

This project is licensed under the MIT License.

See LICENSE.md file for details.