Accurate sprite rotation with mouse movement

Question

I am using the current mouse position to aim the player sprite, but I am not getting the best results.

It will often by off by a noticeable amount, or even a complete ~180 off if I go from one side, through the player, to the other side.

boolean mouseMoved(int screenX, int screenY) {
    mouseX = screenX;mouseY = screenY;
    rot = MathUtils.radiansToDegrees * MathUtils.atan2((float)mouseY - (float)player.y, (float)mouseX - (float)player.x);
    if (rot < 0) rot += 360;
    player.setRot(rot);
}

Update 1:

When I am at 90/270 it seems to pickup the angle fine. But for 0/180, the angle is off depending on how far away my cursor is to a certain point (not the player's point).

Update 2:

Added in semi-fix, but I'm still having the same exact problem

Update 3:

I was mistaken the semi-fix wasn't actually a fix. I fixed it for real, though. The problem was that libgdx has (0,0) at the bottom left, but the InputProcessor has (0,0) at the top left. What I had to do was adjust the mouse position so that it's (0,0) was the bottom left. I did this by

mouseX = screenX;
mouseY = Game.HEIGHT - screen y;

This will give the player coordinates & mouse coordinates the same origin, and then the formula will work.

Answer 1

Try multiplying the y value with -1.

And make sure to convert the mouse and playercoordinates to the world coordinates.

I had this exact problem yesterday. It has to do with wrong conversion - think about the two different coordinate-systems (Viewport- and World-coordinates)

Answer 2

Try using different math techniques for the rotation: Instead of degrees, try using Quaternions. It's the most used rotation technique for 3D objects in professional engines, because degrees don't work as well in 3D.

There's more on Wikipedia and I found some Libgdx code.

I have more materials, but I can't include them due to my low reputation, but you can find lots on Google.

Answer 3

First, inverse the player position and the mouse position. This way, you will have the real angle from the player to the mouse, and not the opposite.

Second, atan2 takes the y difference as the first argument, and the x difference as the second. That is probably why your code looks like it fails.

Your code should look like this with the 2 corrections I mentioned:

boolean mouseMoved(int screenX, int screenY) {
    mouseX = screenX;
    mouseY = screenY;
    // Inverted mouse and player position as well as X and Y axis
    float angle = (float)Math.toDegrees(Math.atan2(mouseY - player.y, mouseX - player.x));
    players.setRot(angle);
}

Source: Experience. Here's a link demonstrating the X and Y axis in the atan2 function (notice how Y comes first).

Answer 4

The problem was that libgdx has the origin (0,0) in the bottom left corner, but the mouse movement from the InputProcessor has (0,0) in the top left. What I had to do was adjust the mouse y position so that its (0,0) was the bottom left. I did this with:

mouseX = screenX;
mouseY = Game.HEIGHT - screenY;

This will give the player coordinates and mouse coordinates the same origin, and then the formula will work.

Here is the entire working code snippet:

boolean mouseMoved(int screenX, int screenY) {
    mouseX = screenX;
    mouseY = Game.HEIGHT - screenY;
    rot = MathUtils.radiansToDegrees * MathUtils.atan2(mouseY - player.y, mouseX - player.x);
    player.setRot(rot);
}