Third-Person Planet Exploration with Three.js

I should share my experiments more often on this blog, and maybe not only the 3D javascript game development that I’m just having fun with. But anyway, here is the last one in date : a third person planet exloration thing.

It’s not exactly “work in progress” since I don’t really plan on improving any of it from here, so just consider this work bare and unfinished. You can try it out here : battle-royale.webmaestro.fr.

3D models of trees

The 3D models used, such as the trees and stones, are from Poly by Google.

The planet is generated when page loads. Noise is applied to the shere vertices length to create the terrain, and “biomes” (materials and 3D models) are set according to the elevation and latitude.

3D Planet from a Noise Sphere

Controls are the classic W, A, S, D and mouse. I had to adapt the “third-person” logic to rotate rather than translate over space.

There is a day and night cycle that depends on where the player is positionned on the globe. The sun and the moon are casting light on opposite sides while turning around.

Day and Night Cycle

The water is… just ugly. There is no collision detection. And the character is a simple cone.

Oh, and there is no server to make it a multi-player shooter, even though that was the ispiration. The idea came when a friend showed me the very entertaining Fortnite. We thought it would be fun to turn this “Battle Royale” island into a planet. Instead of a “storm” shrinking toward the gathered players, we could simply reduce the radius of the spherical terrain… That was the concept.

Maybe I could post details about the code if whoever is interested. In the meantime I have other things to focus on !

3D Noise Sphere Geometry with Three.js

This extended Three.js geometry applies noise elevation over a sphere.

class NoiseSphereGeometry extends THREE.SphereGeometry {
    constructor(radius, widthSegments, heightSegments, {seed, noiseWidth, noiseHeight}) {
        super(radius, widthSegments, heightSegments);
        const getNoise = (vertice) => ImprovedNoise.noise(
                seed + vertice.x / noiseWidth,
                seed + vertice.y / noiseWidth,
                seed + vertice.z / noiseWidth
            ),
            noiseMap = this
                .vertices
                .map(getNoise),
            noiseMax = Math.max(...noiseMap),
            noiseMin = -Math.min(...noiseMap);
        for (const v in this.vertices) {
            if (noiseMap[v] > 0) {
                this
                    .vertices[v]
                    .elevation = noiseMap[v] / noiseMax;
            } else {
                this
                    .vertices[v]
                    .elevation = noiseMap[v] / noiseMin;
            }
            this
                .vertices[v]
                .multiplyScalar(1 + this.vertices[v].elevation * noiseHeight / radius);
        }
    }
}

Make sure to import the ImprovedNoise function from the Three.js examples.

<script src="three/examples/js/ImprovedNoise.js"></script>

Basic Collisions Detection, RayCasting with Three.Js

I thought I would have to use a physics engine (like Cannon.Js or Ammo.Js), but Three.Js on its own is enough to sort us out with collisions, thanks to its Raycaster’s .intersectObjects() method.

This post follows a previous one, about setting up a scene and basic character controls with Three.Js, that maybe you should read if you need to understand a bit more those Character and World classes I’m playing with here.

Using the RayCaster

Even after a little read about raycasting, I’m not quite sure I can define its concept properly… But let’s put it like this : from one origin (the very center position of our character’s Object3D), we’re going to spread vectors in every direction we’re able to move. For each one of those “rays“, we’ll be able to test if it intersects with any given mesh, and if so, to disable any move in that direction.

Collecting the obstacles

So first of all, we need to collect every obstacle in an array : all the meshes that we’re not supposed to cross.

var World = Class.extend({
  /* ... */
  getObstacles: function () {
    'use strict';
    return this.obstacles.concat(this.walls);
  }
});

Testing and prevent collisions

Our little character motions are based on its .direction vector. Now by testing every possible direction (with the rays), we’ll be able to update that vector for it not to drive us into an obstacle.

var Character = Class.extend({
  // Class constructor
  init: function (args) {
    /* ... */
    // Set the character modelisation object
    this.mesh = new THREE.Object3D();
    /* ... */
    // Set the rays : one vector for every potential direction
    this.rays = [
      new THREE.Vector3(0, 0, 1),
      new THREE.Vector3(1, 0, 1),
      new THREE.Vector3(1, 0, 0),
      new THREE.Vector3(1, 0, -1),
      new THREE.Vector3(0, 0, -1),
      new THREE.Vector3(-1, 0, -1),
      new THREE.Vector3(-1, 0, 0),
      new THREE.Vector3(-1, 0, 1)
    ];
    // And the "RayCaster", able to test for intersections
    this.caster = new THREE.Raycaster();
  },
  // Test and avoid collisions
  collision: function () {
    'use strict';
    var collisions, i,
      // Maximum distance from the origin before we consider collision
      distance = 32,
      // Get the obstacles array from our world
      obstacles = basicScene.world.getObstacles();
    // For each ray
    for (i = 0; i < this.rays.length; i += 1) {
      // We reset the raycaster to this direction
      this.caster.set(this.mesh.position, this.rays[i]);
      // Test if we intersect with any obstacle mesh
      collisions = this.caster.intersectObjects(obstacles);
      // And disable that direction if we do
      if (collisions.length > 0 && collisions[0].distance <= distance) {
        // Yep, this.rays[i] gives us : 0 => up, 1 => up-left, 2 => left, ...
        if ((i === 0 || i === 1 || i === 7) && this.direction.z === 1) {
          this.direction.setZ(0);
        } else if ((i === 3 || i === 4 || i === 5) && this.direction.z === -1) {
          this.direction.setZ(0);
        }
        if ((i === 1 || i === 2 || i === 3) && this.direction.x === 1) {
          this.direction.setX(0);
        } else if ((i === 5 || i === 6 || i === 7) && this.direction.x === -1) {
          this.direction.setX(0);
        }
      }
    }
  },
  // Process the character motions
  motion: function () {
    'use strict';
    // Update the directions if we intersect with an obstacle
    this.collision();
    // If we're not static
    if (this.direction.x !== 0 || this.direction.z !== 0) {
      // Rotate the character
      this.rotate();
      // Move the character
      this.move();
      return true;
    }
  }
  /* ... */
});

And that’s it. So of course this isn’t perfect : some meshes are not taken into account (the hands and feet for instance), and it’s possible for some obstacles to get through our rays (try to walk just a little bit too close beside the cube on the demo). One solution would be to add some more rays. But still, this is meant to stay basic, so I’ll keep it like this so far.

See ya folks !

Basic Character Controls with Three.Js

I wanted to try Three.Js out since a little while, and I finally found some time for it. I’ll detail here the different classes I wrote for this to work.

Notice that the collisions are not handled yet. But this will be the topic of an upcoming post, as soon as I figure it out !

I used John Resig’s Simple JavaScript Inheritance in order to get rid of those ugly prototype declarations, to get a cleaner code, and to simplify any further implementation that would require classes inherithance.

Oh, that, and Paul Irish’s requestAnimationFrame polyfill.

That’s it for the tools. You don’t have to use them, but I will in the following explanations. In case you don’t, just replace the classes declarations by their prototype equivalences.

Now the first thing to say about what I experimented : so much trigonometry fun ! Let’s go.

Set the scene

I’m rendering within a <figure> tag.

<figure id="basic-scene"></figure>

Now we got to create the scene. We need a camera, to define from which view point we’re looking, some light or we won’t see a thing, and a “renderer”, which is actually the canvas viewport we render with.

This main class here will also set some event listeners for us to handle the user’s interactions, and a “frame” method that we’ll call at every animation frame request.

var basicScene;
var BasicScene = Class.extend({
    // Class constructor
    init: function () {
        'use strict';
        // Create a scene, a camera, a light and a WebGL renderer with Three.JS
        this.scene = new THREE.Scene();
        this.camera = new THREE.PerspectiveCamera(45, 1, 0.1, 10000);
        this.scene.add(this.camera);
        this.light = new THREE.PointLight();
        this.light.position.set(-256, 256, -256);
        this.scene.add(this.light);
        this.renderer = new THREE.WebGLRenderer();
        // Define the container for the renderer
        this.container = jQuery('#basic-scene');
        // Create the user's character
        this.user = new Character({
            color: 0x7A43B6
        });
        this.scene.add(this.user.mesh);
        // Create the "world" : a 3D representation of the place we'll be putting our character in
        this.world = new World({
            color: 0xF5F5F5
        });
        this.scene.add(this.world.mesh);
        // Define the size of the renderer
        this.setAspect();
        // Insert the renderer in the container
        this.container.prepend(this.renderer.domElement);
        // Set the camera to look at our user's character
        this.setFocus(this.user.mesh);
        // Start the events handlers
        this.setControls();
    },
    // Event handlers
    setControls: function () {
        'use strict';
        // Within jQuery's methods, we won't be able to access "this"
        var user = this.user,
            // State of the different controls
            controls = {
                left: false,
                up: false,
                right: false,
                down: false
            };
        // When the user presses a key 
        jQuery(document).keydown(function (e) {
            var prevent = true;
            // Update the state of the attached control to "true"
            switch (e.keyCode) {
                case 37:
                    controls.left = true;
                    break;
                case 38:
                    controls.up = true;
                    break;
                case 39:
                    controls.right = true;
                    break;
                case 40:
                    controls.down = true;
                    break;
                default:
                    prevent = false;
            }
            // Avoid the browser to react unexpectedly
            if (prevent) {
                e.preventDefault();
            } else {
                return;
            }
            // Update the character's direction
            user.setDirection(controls);
        });
        // When the user releases a key
        jQuery(document).keyup(function (e) {
            var prevent = true;
            // Update the state of the attached control to "false"
            switch (e.keyCode) {
                case 37:
                    controls.left = false;
                    break;
                case 38:
                    controls.up = false;
                    break;
                case 39:
                    controls.right = false;
                    break;
                case 40:
                    controls.down = false;
                    break;
                default:
                    prevent = false;
            }
            // Avoid the browser to react unexpectedly
            if (prevent) {
                e.preventDefault();
            } else {
                return;
            }
            // Update the character's direction
            user.setDirection(controls);
        });
        // On resize
        jQuery(window).resize(function () {
            // Redefine the size of the renderer
            basicScene.setAspect();
        });
    },
    // Defining the renderer's size
    setAspect: function () {
        'use strict';
        // Fit the container's full width
        var w = this.container.width(),
            // Fit the initial visible area's height
            h = jQuery(window).height() - this.container.offset().top - 20;
        // Update the renderer and the camera
        this.renderer.setSize(w, h);
        this.camera.aspect = w / h;
        this.camera.updateProjectionMatrix();
    },
    // Updating the camera to follow and look at a given Object3D / Mesh
    setFocus: function (object) {
        'use strict';
        this.camera.position.set(object.position.x, object.position.y + 128, object.position.z - 256);
        this.camera.lookAt(object.position);
    },
    // Update and draw the scene
    frame: function () {
        'use strict';
        // Run a new step of the user's motions
        this.user.motion();
        // Set the camera to look at our user's character
        this.setFocus(this.user.mesh);
        // And draw !
        this.renderer.render(this.scene, this.camera);
    }
});

Hello World

Or so… Because of obvious performance reasons, we won’t recreate an exact 3D representation of planet earth. We’ll create one ground and four walls instead. But I’ll call it world anyway.

Using a Object3D instead of a simple Mesh allows us to update independently the geometries within this very group of meshes.

var World = Class.extend({
    // Class constructor
    init: function (args) {
        'use strict';
        // Set the different geometries composing the room
        var ground = new THREE.PlaneGeometry(512, 1024),
            height = 128,
            walls = [
                    new THREE.PlaneGeometry(ground.height, height),
                    new THREE.PlaneGeometry(ground.width, height),
                    new THREE.PlaneGeometry(ground.height, height),
                    new THREE.PlaneGeometry(ground.width, height)
            ],
            obstacles = [
                    new THREE.CubeGeometry(64, 64, 64)
            ],
            // Set the material, the "skin"
            material = new THREE.MeshLambertMaterial(args),
            i;
        // Set the "world" modelisation object
        this.mesh = new THREE.Object3D();
        // Set and add the ground
        this.ground = new THREE.Mesh(ground, material);
        this.ground.rotation.x = -Math.PI / 2;
        this.mesh.add(this.ground);
        // Set and add the walls
        this.walls = [];
        for (i = 0; i < walls.length; i += 1) {
            this.walls[i] = new THREE.Mesh(walls[i], material);
            this.walls[i].position.y = height / 2;
            this.mesh.add(this.walls[i]);
        }
        this.walls[0].rotation.y = -Math.PI / 2;
        this.walls[0].position.x = ground.width / 2;
        this.walls[1].rotation.y = Math.PI;
        this.walls[1].position.z = ground.height / 2;
        this.walls[2].rotation.y = Math.PI / 2;
        this.walls[2].position.x = -ground.width / 2;
        this.walls[3].position.z = -ground.height / 2;
        // Set and add the obstacles
        this.obstacles = [];
        for (i = 0; i < obstacles.length; i += 1) {
            this.obstacles[i] = new THREE.Mesh(obstacles[i], material);
            this.mesh.add(this.obstacles[i]);
        }
        this.obstacles[0].position.set(0, 32, 128);
    }
});

And now the funny part…

Live little 3D character, live !

Here as well, we use an Object3D in order to group our meshes.

Our character has two different types of motion to update : its position and its rotation. Further on, we could easily make it jump, dive, or dance. But let’s start with a simple “moving around” action only.

A direction vector will represent the motion that our user is calling through the controls.

The step property will record the progression of the character’s position motion. We will use it to animate its feet and hands.

The feet of our simple character are half-spheres. The trick is simply to properly set the phiStart, phiLength, thetaStart and thetaLength parameters of the SphereGeometry method.

var Character = Class.extend({
    // Class constructor
    init: function (args) {
        'use strict';
        // Set the different geometries composing the humanoid
        var head = new THREE.SphereGeometry(32, 8, 8),
            hand = new THREE.SphereGeometry(8, 4, 4),
            foot = new THREE.SphereGeometry(16, 4, 4, 0, Math.PI * 2, 0, Math.PI / 2),
            nose = new THREE.SphereGeometry(4, 4, 4),
            // Set the material, the "skin"
            material = new THREE.MeshLambertMaterial(args);
        // Set the character modelisation object
        this.mesh = new THREE.Object3D();
        this.mesh.position.y = 48;
        // Set and add its head
        this.head = new THREE.Mesh(head, material);
        this.head.position.y = 0;
        this.mesh.add(this.head);
        // Set and add its hands
        this.hands = {
            left: new THREE.Mesh(hand, material),
            right: new THREE.Mesh(hand, material)
        };
        this.hands.left.position.x = -40;
        this.hands.left.position.y = -8;
        this.hands.right.position.x = 40;
        this.hands.right.position.y = -8;
        this.mesh.add(this.hands.left);
        this.mesh.add(this.hands.right);
        // Set and add its feet
        this.feet = {
            left: new THREE.Mesh(foot, material),
            right: new THREE.Mesh(foot, material)
        };
        this.feet.left.position.x = -20;
        this.feet.left.position.y = -48;
        this.feet.left.rotation.y = Math.PI / 4;
        this.feet.right.position.x = 20;
        this.feet.right.position.y = -48;
        this.feet.right.rotation.y = Math.PI / 4;
        this.mesh.add(this.feet.left);
        this.mesh.add(this.feet.right);
        // Set and add its nose
        this.nose = new THREE.Mesh(nose, material);
        this.nose.position.y = 0;
        this.nose.position.z = 32;
        this.mesh.add(this.nose);
        // Set the vector of the current motion
        this.direction = new THREE.Vector3(0, 0, 0);
        // Set the current animation step
        this.step = 0;
    },
    // Update the direction of the current motion
    setDirection: function (controls) {
        'use strict';
        // Either left or right, and either up or down (no jump or dive (on the Y axis), so far ...)
        var x = controls.left ? 1 : controls.right ? -1 : 0,
            y = 0,
            z = controls.up ? 1 : controls.down ? -1 : 0;
        this.direction.set(x, y, z);
    },
    // Process the character motions
    motion: function () {
        'use strict';
        // (if any)
        if (this.direction.x !== 0 || this.direction.z !== 0) {
            // Rotate the character
            this.rotate();
            // And, only if we're not colliding with an obstacle or a wall ...
            if (this.collide()) {
                return false;
            }
            // ... we move the character
            this.move();
            return true;
        }
    },
    // Rotate the character
    rotate: function () {
        'use strict';
        // Set the direction's angle, and the difference between it and our Object3D's current rotation
        var angle = Math.atan2(this.direction.x, this.direction.z),
            difference = angle - this.mesh.rotation.y;
        // If we're doing more than a 180°
        if (Math.abs(difference) > Math.PI) {
            // We proceed to a direct 360° rotation in the opposite way
            if (difference > 0) {
                this.mesh.rotation.y += 2 * Math.PI;
            } else {
                this.mesh.rotation.y -= 2 * Math.PI;
            }
            // And we set a new smarter (because shorter) difference
            difference = angle - this.mesh.rotation.y;
            // In short : we make sure not to turn "left" to go "right"
        }
        // Now if we haven't reached our target angle
        if (difference !== 0) {
            // We slightly get closer to it
            this.mesh.rotation.y += difference / 4;
        }
    },
    move: function () {
        'use strict';
        // We update our Object3D's position from our "direction"
        this.mesh.position.x += this.direction.x * ((this.direction.z === 0) ? 4 : Math.sqrt(8));
        this.mesh.position.z += this.direction.z * ((this.direction.x === 0) ? 4 : Math.sqrt(8));
        // Now let's use Sine and Cosine curves, using our "step" property ...
        this.step += 1 / 4;
        // ... to slightly move our feet and hands
        this.feet.left.position.setZ(Math.sin(this.step) * 16);
        this.feet.right.position.setZ(Math.cos(this.step + (Math.PI / 2)) * 16);
        this.hands.left.position.setZ(Math.cos(this.step + (Math.PI / 2)) * 8);
        this.hands.right.position.setZ(Math.sin(this.step) * 8);
    },
    collide: function () {
        'use strict';
        // INSERT SOME MAGIC HERE
        return false;
    }
});

I started to look at Cannon.JS to deal with the collisions, but it will have to wait for another post.

Now let’s call our main object and start the animated rendering.

basicScene = new BasicScene();
function animate () {
    requestAnimationFrame(animate);
    basicScene.frame();
}
animate();

And here we are !

Conclusion

Isn’t it surprising how easy it became to set up in-browser 3D ? Thanks Mr.doob for Three.JS !

This is only an experiment on a basic set up, but I would like to upgrade it every now and then. The next step will be the collisions detection.

Have fun fellows, I hope I’ll see you around again !

Board Game Experiment, with JS Prototypes and Canvas

Okay fellows. It took me ages, but I coded my very first game in JavaScript. It’s based on Quoridor, a really interesting “abstract strategy” board game, designed by Mirko Marchesi and published by Gigamic. The rules and logic of it are quite simple, so I thought it would be a nice project for my first experiment of this kind.

What I’m about to do here, is to review step by step how I made it. It was my first real use of Javascript Prototypes and HTML5 canvas, and I learned a lot, so maybe this little “tutorial” will be interesting for you.

  1. Introduction
  2. Markup
  3. Rules
  4. Classes
  5. Game Object
  6. Rendering
  7. Square Object
  8. requestAnimationFrame
  9. Player Object
  10. User Player
  11. Computer Player
  12. Fences Manipulation
  13. Setting Form
  14. Style Sheet
  15. Launching
The game “Quoridor” is legally owned and protected by Gigamic, so you guys understand that this code can’t go further than our private experimental use.

Maybe you would like to play the demo before we make this work and rock.