3D Game Development Terminology

Before you start, make certain you are familiar with the following concepts and terminology.

3D Graphics and Audio

OpenGL is the Open Graphics Library, a platform-independent specification for rendering 2D/3D computer graphics. For Java, there are two implementations of OpenGL-based renderers:

  1. Lightweight Java Game Library (LWJGL)
  2. Java OpenGL (JOGL)

OpenAL is the Open Audio Library, a platform-independent 3D audio API.

Context, Display, Renderer

The jME Context makes settings, renderer, timer, input and event listeners, display system, accessible to a JME game.

Geometry

Coordinates

Coordinates represent a location in a coordinate system, relative to the origin (0,0,0). m. In 3D space, you need to specify three coordinate values to locate a point: x,y,z. As opposed to a vector (which looks similar), a coordinate does not have a "direction".

The Origin

The origin is the central point in the 3D world. It's at the coordinates (0,0,0). Code sample: Vector3f origin = new Vector3f( Vector3f.ZERO );

Vectors

A vector has a length and a direction. It is used like an arrow pointing at a point in 3D space. A vector starts at the origin (0,0,0), and ends at the target coordinate (x,y,z). Backwards directions are expressed with negative values. Code sample: Vector3f v = new Vector3f( 17 , -4 , 0 );

Unit Vectors

A unit vector is a basic vector with a length of 1 world unit. Since its length is fixed (and it thus can only point at one location anyway), the only interesting thing about this vector is its direction.

Normalized Vectors

A normalized vector is a custom unit vector. A normalized vector is not the same as a (surface) normal vector. When you normalize a vector, it still has the same direction, but you lose the information where the vector originally pointed. For instance, you normalize vectors before calculating angles.

Surface Normals

A surface normal is a vector that is perpendicular to a plane. You calculate the Surface Normal by calculating the cross product.

Cross Product

The cross product is a calculation that you use to find a perpendicular vector (an orthogonal, a "right angle" at 90°). In 3D space, speaking of an orthogonal only makes sense with respect to a plane. You need two vectors to uniquely define a plane. The cross product of the two vectors, v1 × v2, is a new vector that is perpendicular to this plane. A vector perpendicular to a plane is a called Surface Normal. Example: The x unit vector and the y unit vector together define the x/y plane. The vector perpendicular to them is the z axis. JME can calculate that this equation is true:
( Vector3f.UNIT_X.cross( Vector3f.UNIT_Y ) ).equals( Vector3f.UNIT_Z ) == true

Polygon, Mesh, Vertex

Most visible objects in a 3D scene are made up of polygon meshes – characters, terrains, buildings, etc. A mesh is a grid-like structure that represents a complex shape. The advantage of a mesh is that it is mathematically simple enough to render in real time, and detailed enough to be recognizable. Every shape is reduced to a number of connected polygons, usually triangles; even round surfaces such as spheres are reduced to a grid of triangles. The polygons' corner points are called vertices. Every vertex is positioned at a coordinate, all vertices together describe the outline of the shape. You create 3D meshes in tools called mesh editors, e.g in Blender. Learn more about 3D maths here.

Materials: Color, Lighting/Shading

What we call "color" is merely part of an object's light reflection. The onlooker's brain uses shading and reflecting properties to infer an object's shape and material. Factors like these make all the difference between chalk vs milk, skin vs paper, water vs plastic, etc! (External examples)

Color

Ambient color

Diffuse color

Light Sources

Emissive color

Reflections

Shininess

Specular Color

Materials: Textures

Textures are part of Materials. In the simplest case, an object could have just one texture, the Color Map, loaded from one image file. When you think back of old computer games you'll remember this looks quite plain. The more information you (the game designer) provide additionally to the Color Map, the higher the degree of detail and realism. Whether you want photo-realistic rendering or "toon" rendering (Cel Shading), everything depends on the quality of your materials and texture maps. Modern 3D graphics use several layers of information to describe one material, each layer is a Texture Map.

Texture Maps

Color Map

jmonkeyengine.googlecode.com_svn_trunk_engine_src_test-data_textures_terrain_splat_road.jpg

Diffuse Map

jmonkeyengine.googlecode.com_svn_trunk_engine_src_test-data_models_hovertank_tank_diffuse.jpg

Bump Map

Bump maps are used to describe detailed shapes that would be too hard or simply too inefficient to sculpt in a mesh editor. You use Normal Maps to model cracks in walls, rust, skin texture, or a canvas weave. You use Height Maps to model whole terrains with valleys and mountains.

Height Map

Normal Map

Specular Map

jmonkeyengine.googlecode.com_svn_trunk_engine_src_test-data_models_hovertank_tank_specular.jpg

Seamless Tiled Textures

jmonkeyengine.googlecode.com_svn_trunk_engine_src_test-data_textures_terrain_brickwall_brickwall.jpg This is a very simple, commonly used type of texture. When texturing a wide area (e.g. walls, floors), you don't create one huge texture, instead you tile a small texture repeatedly to fill the area. A seamless texture is an image file that has been designed so that it can be used as tiles: The right edge matches the left edge, and the top edge matches the bottom edge. The onlooker cannot easily tell where one starts and the next one ends, thus creating an illusion of a huge texture. The downside is that the tiling becomes painfully obvious when the area is viewed from a distance. Also you cannot use it on more complex models such as characters.

UV Maps

jmonkeyengine.googlecode.com_svn_trunk_engine_src_test-data_models_ferrari_car.jpg Creating a texture for a cube is easy – but what about a character with a face and extremities? For more complex objects, you design the texture in the same ways as a sewing pattern: One image file contains the outline of the front, back, and side of the object, next to one another. Specific areas of the flat texture (UV coordinates) map onto certain areas of your 3D model (XYZ coordinates), hence the name UV map. Getting the seams and mappings right is crucial: You will have to use a graphic tool like Blender to create UV Maps. It's worth the while to learn this, UV mapped models look a lot more professional.

Environment Maps

Environment Maps are used to create the impression of reflections and refractions. You create a Cube Map to represent your environment, similar to a skybox. (Sphere Maps are possible, but often look too distorted.) You give the Cube Map a set of images showing a "360° view" of the completed scene. The renderer uses this as base to texture the reflective surface. Of course these reflections are static and not "real", e.g. the player will not see his avatar's face reflected, etc… But they cause the desired "glass/mirror/water" effect, and are fast enough to render in real usecases, it's better than nothing.

MIP Map Texture

You provide the texture in two or three resolutions to be stored in one file (MIP = "multum in parvo" = "many in one"). Depending on how close (or far) the camera is, the engine automatically renders a more (or less) detailed texture for the object. Thus objects look smooth from close up, but don't waste resources with unspottable details when far away. Good for everything, but requires more time to create and more space to store textures.

Procedural Textures

A procedural texture is generated from repeating one small image, plus some pseudo-random, gradient variations (Perlin noise). Procedural textures look more natural than static rectangular textures, for instance, they look less distorted on spheres. On big meshes, their repetitiveness is much less noticable than tiled seamless textures. Procedural textures are ideal for irregular large-area textures like grass, soil, rock, rust, and walls. See also: jMonkeyPlatform NeoTexture plugin jmonkeyengine.org_wp-content_uploads_2010_10_neotexture-2.jpg See also: Creating Materials in Blender, Blender: Every Material Known to Man

Animation

In 3D games, Skeletal Animation is used for animated characters, but in principle the skeleton approach can be extended to any 3D mesh (for example, an opening crate's hinge can be considered a joint). Unless you animate a 3D cartoon, realism of animated characters is generally a problem: Movement can look alien-like mechanical or broken, the character appears hollow, or as if floating. Professional game designers invest a lot of effort to make characters animate in a natural way (including motion capturing).

Rigging and Skinning

An animated character has an armature: An internal skeleton (Bones) and an external surface (Skin). The Skin is the visible outside of the character and it includes clothing. The Bones are not visible and are used to interpolate (calculate) the morphing steps of the skin. JME3, the game engine, only loads and plays your recorded animations. You must use a tool (such as Blender) to set up (rig, skin, and animate) a character.

  1. Rigging: The Construction of a character's skeleton.
    • Create as few Bones as possible to decrease complexity.
    • Bones are connected in a parent-child hierarchy: Moving one bone can pull another bone with it (e.g. arm pulls hand).
    • Bones follow a certain naming scheme so the 3D engines know what's what.
  2. Skinning: The association of individual bones with the corresponding skin sections.
    • Each Bone is connected to a part of the Skin. Animating the (invisible) Bone pulls the (visible) Skin with it.
      E.g. the thigh Bone is connected to the upper leg Skin.
    • One part of the Skin can be affected by more than one bone (e.g. knee, elbow).
    • The connection between bones and skin sections is gradual: You assign weights how much each skin polygon is affected by any bone's motion.
      E.g. when the thigh bone moves, the leg is fully affected, the hips joints less so, and the head not at all.
  3. Keyframe Animation: A keyframe is one recorded snapshot of a motion sequence.
    • A series of keyframes makes up one animation.
    • Each model can have several animations. Each animation has a name to identify it (e.g. "walk", "attack", "jump").
    • You specify in your game code which keyframe animation to load, and when to play it.

Kinematics

Controller and Channel

In the JME3 application, you register animated models to the Animation Controller. The controller object gives you access to the available animation sequences. The controller has several channels, each channels can run one animation sequence at a time. To run several sequences, you create several channels, and run them in parallel.

Artificial Intelligence (AI)

Non-player (computer-controlled) characters (NPCs) are only fun in a game if they do not stupidly bump into walls, or blindly run into the line of fire. You want to make NPCs "aware" of their surroundings and let them make decisions based on the game state – otherwise the player can just ignore them. The most common use case is that you want to make enemies interact in a way so they offer a more interesting challenge for the player. "Smart" game elements are called artificially intelligent agents (AI agents). An AI agent can be an NPC, but also an "automatic alarm system" that locks doors after an intruder alert, or a trained pet. The domain of artificial intelligence deals, among other things, with:

There are lots of resources explaining interesting AI algorithms:

view online version