Progression on the project

A number of areas have progressed with the project so I thought it was about time I updated my blog.

Firstly the most noticeable update is on the graphics front. Here's a screenshot of the current build:


1) 3-D textures
The terrain is now textured with a 3-D texture. 3-D textures are basically stacks of standard 2-D textures which allows you to gently blend between different layers. In the case of the terrain, there are 6 different layers ranging from sand to dirt, to grass, to smoother grass, to rougher grass and eventually to a snow type texture. In the case of the terrain, the height at each point dictates which texture layer should be used. This means that low areas near the water have a sandy beach effect whilst high areas blend into snow capped mountains.

I noticed that originally the terrain looked very artificial because the layers were very clear, so I randomly offset the layer on each point of the terrain to add extra variation. This isn't the best solution, but it certainly gives an overall more natural feel. I may create a separate blend map layer to allow for extra terrain features such as roads.

2) Shaders
I have implemented a system to use the OpenGL shader language (GLSL) to create some shader effects. Currently this supports per-pixel diffuse lighting on the terrain and models. I want to extend this for more complex materials, bump mapping and special effects.

3) OBJ Model loading
I have chosen to use the OBJ format for model loading as nearly every 3-D modelling program supports it and its a relatively easy format to import. The model loader will read position, normal and texture coordinate data, and will generate normal data if none is available. It also generates edge data for use in shadow volumes (see below)

4) Shadows
I have opted to use a shadow volume technique for generating shadows. Basically for each object a volume is created to represent the space which is occluded from the light source. The area where this volume intersects with surrounding geometry is where it is in shadow. I'm not quite happy with my current implementation as there are a few graphical glitches. Also the lighting model is not quite finished which is why the shadows are completely blacked out. I'll post more on the topic when the lighting model is in place.

5) Object instancing
I have written a dynamically sized object container which allows for speedy random access. It is basically a tree structure which increases in size as you add more objects. Each object is given an ID value which means searching for it is just a matter of traversing the correct nodes on the tree. This is a big plus as it means when it comes to implementing multiplayer, objects can be referenced by ID during communication.

6) Picking
I have implemented a system to decipher which object in the scene the mouse pointer is hovering over which will be very useful for the overall input system. All objects are redrawn in the scene in 'select' mode with each object assigned an ID value. The same system should allow the user to click and drag a rectangular area to select multiple objects.

Thats about it for now. Here are some of the next things I want to be working on:
- Lighting model (ambient occlusion, directional light source, correct shadowing)
- Selecting objects
+- Object messaging infrastructure (issuing commands)
+- Movement and pathfinding
- Animation
- Occlusion culling

Terrain rendering

This is an early preview of the terrain rendering engine. It's using a heightmap generated from Terragen which is then imported as a grey scale bitmap. The program is rendering the mesh without any material textures but does use a precalculated ambient map.

The ambient light map is calculated by using a ray-casting algorithm to estimate the angle of the horizon at various angles for each vertex on the height field. Using these angle values, an approximation of the sky lighting contribution can be made.

The next step for lighting is to calculate the sun contribution. Using the same ray-casting algorithm the angle of the horizon will be stored in a horizon map. This can then be compared with the angle of the sun to determine if a point is in shadow or in direct sun light. The added benefit of using a horizon map is that the angle of the sun can be changed with light and shadow being generated dynamically, so different times of day can be simulated.

Pixel shaders are going to be used to incorporate the lighting and materials. I hope to also support bump mapping to give the terrain a more rough 'feel' to it.

Change of project direction

After some careful consideration and weighing up the advantages and disadvantages, I have decided to change the project so it will no longer be a Java applet embedded in a webpage, and will instead be an offline application that can be downloaded and played single player or multiplayer, internet or LAN. The game will now be written in C++ and use OpenGL for the graphics rendering.

The main reason for this change is that I was finding Java to be quite a limitation and although this was pushing me to write well optimised code, I was unhappy with the quality of the graphics. Graphics is certainly the strongest aspect of my programming skills and it would seem a shame to limit myself so much in this area of the game. The game will now feature a fully 3-D environment making use of available graphics hardware. Lighting and effects will be used by writing custom vertex and pixel shaders.

The game will still remain cross platform, and will be released for Windows, Mac OS X and Linux. I will be using the SDL library to handle input, sound, window management so the porting to other platforms should be fairly trivial.

Concept soldier

I sketched a soldier as part of the concept art for warring states as shown above. I'm now going about making it into a 3-D model for the game.

On some other notes, I've set up a subversion repository for version control and the code is progressing quite nicely. The basic rendering system is in place with an isometric style map being drawn and culled correctly, and sprites in the game are sorted and displayed correctly.

Interesting youtube video

I stumbled across this youtube video which has some really interesting background about the weapons used during the warring states period in china. It will be a very good reference to make when designing the units in the game:

First preview of assets

Here are some of the model designs that I have been working on for the game. The models will all be pre-rendered and used as sprites in the game. As can be seen I haven't added textures or materials yet but the structures are taking shape.

Blog started!

I've started this blog to record the development of my final year university project. It will be an online multiplayer real time strategy (RTS) game which will be played through a web browser. The game will be written in Java and aims to feature multiplayer as well as a challenging computer AI as a single player option.

The name 'Warring States' refers to the period of time that the game will be set. This was a time in China's history where the country was split into seven seperate states, each at war for the control of the known world. In multiplayer games, each players will choose which faction they want to play as, and the single player game will feature a campaign mode which will follow the state of Qin who aim to unite the country.

I hope to update this blog whenever I implement anything particularly interesting. I plan to release a public beta for testing when the game reaches a stable state.