news

Apologies for the lack of posting lately to those who are still visiting. I've started a new project! It is a minimalistic platformer in XNA. Here are some mockups:



It is very Knytt Stories-inspired. I'm aiming for online co-op for two players with somewhat experimental gameplay (more details on that once it is better defined).

The physics based platformer game project is now on hold. It was a difficult project to work on and I want to do something that has a better chance of being completed. I still want to make that game, so I will likely pick it up again later.

Posting will probably be light until the end of the semester. These are my goals for the next few months:

• work out the design of the XNA platformer and get some serious coding done
  
• post some the school work that I did last semester
  
• update the flash game I made for school last semester and make it public
  
• finally make an update to Polyworks

flash platformer physics

I've been experimenting with platformer physics using the Flash version of Box2D in order to familiarize myself with ActionScript. I gave the character the ability to hang on edges and climb ropes by using joints in ways they were never meant to be used.

click to activate and use the arrow keys to control the character

yet more physics animation

Here is a better example of a physics-based walk:



It's still a bit crude but it's approaching something that could be used in the game. The animation system controls the rotation of the box that the feet are attached to, the feet being IK handles that control the animation of the legs. The walking animations are played when the appropriate keys are pressed.

In other news, I'm starting to see the true enormity of this project, and scaling my plans down accordingly by stripping them of unnecessary features. Fortunately, I hadn't posted about those features so there is nobody to disappoint except myself. I think this may strengthen the game though, by removing distractions that could obscure its core.

more physics animation

I decided to try integrating physics right into the character editor. I'm not completely sure yet if it's going to work the way I expect it to, but the plan is to have both the skeletal animation and the physical movement of the character controlled through the animation system.

For example, consider a character that is made up of joints and quads and attached to a box which is connected by a revolute joint to a circle (as in the screenshots). The walking animation would animate the joints to make it look like the character is walking and it would also animate the motor of the revolute joint to rotate the circle so that the character moves along the ground.





I'm still working out the complexities but I think this system has a lot of potential for awesomeness. It is basically a way of doing data-driven character movement, which gives character creators a lot of freedom without having to write any code.

more concept art

I'm thinking about going with a more abstract style than the textured/painted look I was originally planning. The graphics would be much easier to make - I could possibly even do all the artwork for an entire game by myself.





The sprites look to be a little too detailed for the environment. Flat colours might be more suitable there instead of the shading.

animation system

I don't feel like making a new screenshot so here are some tulips instead.

This is a general overview of the way my character editor works.

To recap from my last post about aiming, all of the objects in the editor are nodes in a hierarchy. They are derived from a base class that I call TransformNode. The specialized nodes that I have so far are the Joint, IKHandle, and Quad. Joints are nodes that form a skeleton and can be affected by IK, IKHandles are nodes that control an IK chain (of joints), and Quads are the textured quads that make up the visible part of an entity.

All nodes contain a collection of attributes that are used in animation. The base node has translation and rotation attributes, and specialized nodes can add other attributes to the collection. The Quad, for example, has additional attributes for width, height, and texture coordinates. (This means that texture/sprite animation can be used in combination with skeletal animation.) New attributes are animateable as soon as they are added to the node and exposed to the interface (via a public property with get/set accessors so that they show up in the property grid).

Attributes are animated by setting keyframes in the animation curve that is associated with that attribute. A curve contains a set of keyframes and a reference to the node it affects, along with the attribute index. Keyframes contain data for control points on the animation curve and their tangents. The positions of the points on the curve are the frame of the animation (x) and the value of the attribute at that frame (y). Bezier interpolation is used to interpolate the values between keyframes. An animation is simply a set of curves.

Aim animations have an additional bit of data called an AimVector. It has minimum and maximum rotation values and is attached to a node in the hierarchy. Aim animations are like regular animations except that there are only two keyframes. When aiming, the angle of the vector between the aim vector node position and the target position is compared to the minimum and maximum angles to get an interpolation value. This value is used to do a linear interpolation between the two frames of the aim animation.

When exporting animations, the curve data is interpolated and values are saved for every frame in the animation (so that interpolation does not need to be performed in game).

This editor has evolved a lot over the time I've been working on it. I'm finally satisfied with this particular implementation, although a lot of the functionality was copied from Maya.

animated gif creation

In my previous post, I included an animated gif showing my editor in action. I thought a gif was appropriate in this situation because it was a short animation, uploading it to YouTube caused serious distortions, Photobucket doesn't allow me to host swf files, and because the limited colours make it suitable for gif compression. I used several free tools to make this gif, and thought I'd describe the process in case others find it useful.

First I used CamStudio to record the demonstration. It's a really simple app that outputs avi files and has the option of converting them to swf files.

After that I used RAD Video Tools to convert the avi file into a series of gif files. I set "Adjust (adds/removes frames) to" to 15, so the output would be 15 frames per second of video.

Then I used GIMP to turn the gif files into a single gif animation using the instructions I found here. Like the post says, I opened the first file and imported the rest as layers using File -> Open As Layers. I saved the file as an animation that loops forever with a 60 ms delay between frames. I also set "frame disposal where unspecified" to one frame per layer, and checked "use delay entered above for all frames" and "use disposal entered above for all frames".

The result was a 523 KB file with 266 frames at 272x248 pixels, 17 seconds at about 15 fps.

more aiming

I came up with a better way to do weapon aiming. In order to implement it I had to rewrite some parts of the editor to make it work more like Maya. All of the objects in the editor (Joints, IKHandles, and Quads) are now derived from the same base class, called a TransformNode, which has translation and rotation values. They are all contained in the same hierarchy, with a single root node. Here is an example of how this system can be used for aiming:



In this example, the aim animation rotates the node that the IK handles are grouped to. The shooting/recoil animation moves the IK handles.

It seems so obvious now that this is the correct way to do it; I'm not sure why I didn't think to do it this way earlier. This system uses the IK handles directly, so the hands are always placed correctly. It's also more flexible and consistent than what I had in my previous post.

game project

I figured I should start documenting the development of my game more thoroughly. I've avoided posting my plans because I don't feel certain about them, but there are some things that are not likely to change so I can at least talk about that.

As you can probably tell from previous posts about this project, it's a side-scrolling game inspired by Soldat. I was motivated to start it by the demand from the Soldat community for more modding possibilities in Soldat. I'm borrowing heavily from Soldat in several aspects (2D skeletal animation, terrain made of textured polygons, combat with projectile weapons, mouse aiming) but the similarities mostly end there.

My game is going to be a single-player Metroidvania-style platformer with a focus on exploration and the use of a variety of weapons and items to defeat monsters and progress through the world. I plan to incorporate elements inspired by the Zelda series, especially Wind Waker, into the gameplay and visual style. Like Zelda, it will have a fantasy theme, with possibly some steampunk thrown in (see Chrono Trigger). Features include vehicles, water, and dynamic lighting.

One of my main goals with this project is to make content creation relatively easy and fast. To that end, I'm developing powerful tools for use with the engine (the character editor and the map/general editor that I will start re-writing soon). I'm aiming to have the tools and engine ready to use by September 2009, with a demo ready by May 2010 for use in my portfolio. At that point, assuming I get there, I will decide if it is worth turning into a full game (possibly with online play). However this schedule is not fixed.

I'm using C# with MDX, and I intend to port it to XNA 3.0 eventually.

In upcoming posts I will go into more detail on the internal workings of the game and character editor and on my plans for the game itself.

box2dx

I've decided to use Box2DX, a C# port of Box2D, for physics.



It was easy to integrate into my program and start using it, and it should save me a lot of trouble. Writing my own code for rigid body dynamics would have taken me forever. It also opens up more potential for physics-based gameplay.

weapon aiming

I got my aiming animation stuff working, sort of. The editor now allows one animation to be added on top of another. This allows me to blend a shooting animation with an aim animation, like this crude example shows:



However it doesn't always give good results. My test animation looks alright when aiming upwards and straight ahead, but not so well when aiming downwards:



As you can see, the support hand doesn't stay where it's supposed to. This could be solved by changing the shooting animation or by using IK for the support hand. In the game, the aim animation would be controlled by the crosshair position.

what's taking so long?

You may be wondering why the progress on my project has been so slow. The main reason for this is that as usual, I bit off more than I can chew. It's a much larger project than anything I've done before, and it's somewhat beyond my abilities - that's what makes it such a good learning experience.

I'm not making just a game, but a game engine and a set of tools to go with it. This means that for every feature I want to implement, I have to generalize it. For example, with the weapon aiming system I've been working on, it's necessary to allow any character made in the character editor to be able to aim, and each character will have a different set of weapons that they are able to use. This makes the problem much more difficult to solve than if it only applied to the main character, for example - then it could all be hard coded. I also try to implement each feature that is exposed in the interface in a way that's easy to use when creating content.

When it comes to programming, I look for solutions that are simple and keep the level of complexity low so that it doesn't turn into a maintenance nightmare in the future. In fact I have a hard time getting any code written if I'm not satisfied with the plan I've come up with. This means that I spend a lot of time working stuff out on paper, but it also means that when I do get around to coding, it often works the first time and cuts down on the time I spend debugging.

All of this, plus difficulty concentrating, results in really slow progress. I'm enjoying the process though, and I'll keep chewing on it as long as it's fun.


P.S. I now have access to a camera, so expect more unrelated photos with my posts.

update

According to the earliest timestamp I could find on my game-related files, I started working on it in May of 2006 - it's been two years. I've learned a lot, even though much of the work I've put into it hasn't produced tangible results.

All I have to show for now is a rough implementation of the graph editor:



Currently I'm working on having joints be controlled by the crosshair position, to make it possible for characters to aim their weapons in a convincing way.

bezier interpolation

I just got my Bezier interpolation working:



This thread was very helpful, and I stole the code for the cubic equation solver from here. The code to draw Bezier curves is from here.

Now I have to apply it to my animation curves.

lack of progress report

yay for Bezier curves
Despite having been out of school for two weeks, I have not managed to make much progress on my project. I did a lot of thinking about my animation system though, and I've decided that I will try to implement a simple version of Maya's graph editor.

My original plan to do frame by frame animation with linear interpolation between frames doesn't seem good enough after animating with Maya. I need the smoothness and slow in/slow out that gives the movement weight and makes it convincing. Also, the ability to set keyframes at any point in the animation, rather than regular intervals, will help with the timing.

Hopefully I'll have something interesting to show before I have to go back to school in two more weeks.

inverse kinematics

After discovering the Joint and IK Handle tools in Maya, I decided to implement IK in my animation editor:



It works much better for animation than the particles and constraints; it's like a cross between that system and the bone hierarchy in the old version. I used Cyclic Coordinate Descent for the IK solver.

editor rewrite

The fourth version of my character/model/animation editor, in progress:



I got rid of the bone hierarchy and cleaned up the interface; it's easier and more fun to use now. I've been working with Maya for a few weeks at school so I ended up stealing some interface ideas and I have plans to incorporate more Maya-like features when I re-implement the animation stuff.

side scroller physics

I was working on porting my collision detection stuff from C++ to C#:



It's a simple AABB colliding with line segments using SAT for collision detection and response. I added friction so it's not sliding around everywhere, but this kind of physics doesn't seem appropriate for this game. I'm going to try to make it work more like old side scrolling SNES games, where the movement is more predictable and easier to control.

Also I made a little particle physics test. It's surprisingly easy to implement this kind of system:

lighting implemented

In addition to the light and shadow generation, I have also implemented multiple textures and four polygon layers (lit non-colliding background, lit colliding foreground, unlit colliding foreground, and unlit non-colliding foreground). I've been able to salvage some code and design from the discontinued PW2 but this map editor is still in the very early stages.

lighting

Originally I was going to use 3D polygons and DirectX lights, but I changed my mind because 1) working in real 3D, even in a limited way, is a pain, and 2) it's hard to predict the effects of a light on 3D geometry. Instead I'm going to implement a two dimensional method that includes shadows. I used one of VirtualTT's maps to do some testing in Photoshop.

unlit:


red, green, and blue lights overlapping, night:


shadows and light-blocking polygons, sunset: