CODEWITHSUNDEEP
androidopengl-esopengl-es-2.0
Burf2000
Burf2000

Reputation: 5193

Android : OpenGL 2.0 first person camera

So I am trying to learn OpenGL 2.0 on Android, I did play quite a bit with OpenGL 1 on iOS and really enjoyed it.

My simple question is about the camera and making a 3D environment where you can move around (First person)

Should I be using 

Matrix.setLookAtM(mViewMatrix, 0, eyeX, eyeY, eyeZ, lookX, lookY, lookZ, upX, upY, upZ);

to control the camera and where I am in the world (updated on onDrawFrame) , or setting that on the onSurfaceCreated (once) and using

Matrix.setIdentityM(mViewMatrix, 0);
Matrix.translateM(mViewMatrix, 0, mMoveY, 0.0f, mMoveX);
Matrix.rotateM(mViewMatrix, 0, mDeltaY, 1.0f, 0.0f, 0.0f);
Matrix.rotateM(mViewMatrix, 0, mDeltaX, 0.0f, 1.0f, 0.0f);

instead which feels like I am rotating the world around me.

I have seen examples where they do either, on OpenGL 1 I used to use GLLookAt

Upvotes: 2

Views: 2014

Answers (1)

Matic Oblak
Matic Oblak

Reputation: 16794

Any of the two methods is fine since you can get same results. The general difference is about how you want to store your objects state. For a 3D environment I would always use 3 vectors to determine the object state (position, forward, up) and use modified version of lookAt and modelMatrix that can place the object with same parameters as lookAt. The upside of this approach is that you can directly place the parameters depending on other object, for instance: A guided missile is following you and is always turned towards you no mater where you are or how you move. Then its forward vector is simply taregetPosition-missilePosition (usually normalized). On the other hand if you have to compute the angles you have quite some work, directly asin, acos and a few if statements for each of the 2 angles. Next for instance simple moving around the room, going forward: If you use base vectors, then position = position+forward*speedFactor while with angles you again have to compute what way are you facing and then do the same... (there are quite a few situations where that is useful)

But there are downsides. You need to have your own system to move and rotate those vectors. For instance if you want to say turn to your left for 45 degrees it would look something like this:

forward = (forward+cross(up,forward)*tan(45)).normalized

and this only works for angle in interval (-90, 90). It gets quite the same when turning up but you need to also correct the up vector.

So to wrap it up, IF you create all the methods to work with base vectors (rotations, look at, model matrix...) they are a real labor saving method. But it simply depends on the project you are writing to decide what to use.

Upvotes: 1

Related Questions