Spine+Box2d+Skinned Mesh

ashwani

Hi,
I m new in Using Spine. There's been a few posts about similar questions, but none completely address my question:
I m able to use box2d in Spine with Region Attachment..

*code

AtlasAttachmentLoader atlasLoader = new AtlasAttachmentLoader(atlas){
      public RegionAttachment newRegionAttachment (Skin skin, String name, String path) {
         AtlasRegion region = atlas.findRegion(path);
         Box2dAttachment attachment = new Box2dAttachment(name);
         if (region == null)
            throw new RuntimeException("Region not found in atlas: " + path + " (region attachment: " + name + ")");
         attachment.setRegion(region);
         return attachment;
      }};



for (Slot slot : skeleton.getSlots()) {
         if (!(slot.getAttachment() instanceof Box2dAttachment))continue;
         Box2dAttachment attachment = (Box2dAttachment)slot.getAttachment();

     PolygonShape boxPoly = new PolygonShape();
     boxPoly.setAsBox(attachment.getWidth() / 2 * attachment.getScaleX(),
        attachment.getHeight() / 2 * attachment.getScaleY(), vector.set(attachment.getX(), attachment.getY()),
        attachment.getRotation() * MathUtils.degRad);

     BodyDef boxBodyDef = new BodyDef();
     boxBodyDef.type = BodyType.DynamicBody;
     attachment.body = world.createBody(boxBodyDef);
     attachment.body.setUserData(ud);
     attachment.body.createFixture(boxPoly, 1).setUserData(slot.getData().getName());
     boxPoly.dispose();
     
     System.out.println(slot.getData().getName());
  }


and "box2dattachment" calss is this

static class Box2dAttachment extends RegionAttachment {
        Body body;
        public  Box2dAttachment (String name) {
            super(name);
         }
      }

But is there is any way to use box 2d with Skinned mesh Attachment or Weighted Mesh attachment?

Would love some feedback, suggestions, ideas,
Thanks

*Note: English is not my native language so go easy with this

Søren

As far as the code helps I can't help out. But as far as I know Box2D has a default setting for how many vertices your polygons can have set to 8. This can be changed in the settings file for Box2D. Also note that your shapes must be convex.

Is there a reason you want to use the skinned mesh and not just attach a bounding box ?

Pharan

I think Box2D has a few restrictions that may be in your way.
(1) polygon shapes need to be convex
(2) you can't update shapes every frame, they need to be rigid. or need to be heavily rebuilt if the vertices change. So you can't animate meshes/weightedmeshes and expect the polygon shapes to follow.

I'm not sure if this is all still true. You may have to double check Box2D's current features.
Shiu's recommendation on using Bounding Box attachments from Spine may be the way to go.

ashwani

HI Pharan, Hi Shiu

Thanks for the reply.
Fare speaking the time when i asked this question i didn't know about Bounding Box and i m very new in spine.
I just want the collision detection and currently figuring out with the bounding box.
Can u provide any link or example for using bounding box attachments , it will be very useful.

PS:
Box2d still have 8 vertices issue and shape convex issues :S .

Thanks

Nate

At its simplest, you update the SkeletonBounds, then ask it about intersections.
https://github.com/EsotericSoftware/spine-runtimes/blob/master/spine-sfml/example/main.cpp#L117-L119

ashwani

Hi Nate,

As u can see in this pic 2 same character draw in different position having bounding box
But when i change the animation (for 1st character) bounding box can't follow the animation.

What m doing wrong?

      public void render (PolygonSpriteBatch sb, OrthographicCamera camera) 
   {
      camera.update();
      sb.getProjectionMatrix().set(camera.combined);
         renderer.draw(sb, skeleton);
   }
   

   public void debugRender(PolygonSpriteBatch sb, OrthographicCamera camera)
   {
      camera.update();
      debugRenderer.getShapeRenderer().setProjectionMatrix(camera.combined);
      debugRenderer.draw(skeleton); 
      

   }

   public void update(float dt)
   {
      bounds.update(skeleton, true);
      state.update(Gdx.graphics.getDeltaTime());
      state.apply(skeleton); 
      skeleton.updateWorldTransform();
      

   }

on click

            if(blue.getbounds().aabbContainsPoint(MyInput.Down_x,MyInput.Down_y))
     {         
        blue.setAnimation(0, "Bend_Right Hand punch", false);
     }

Nate

SkeletonRendererDebug has its own SkeletonBounds and this is what you see drawn. It updates the bounds and then draws, so should be correct. There must be something else going on in your app. Does it work as expected in SkeletonViewer? Debug it: simplify your app, verify each of your assumptions.

ashwani

HI Nate

Everything works well in skeleton viewer except bounding boxes

1 doubt : can we have to set bounding in every animation when creating a project in spine??

According to my designer he set boxes only in the default frame.

If we have to set bounding boxes in every frame then can u provide a link for creating a spine with bounding boxes

Thanks :nerd:

Nate

I'm guessing your bounding boxes are all attached to the root bone. You need to attach each one to the appropriate bone.

ashwani

Hi Nate,
yes ur guessing was right.. Works fine now 🙂

I can get the collision detection of two spine body by

                    blue.getbounds().update(blue.getSkeleton(),true);
		orange.getbounds().update(orange.getSkeleton(),true);
		if(blue.getbounds().aabbIntersectsSkeleton(orange.getbounds()))
                     {
                         system.out.println("hit received");
                     }

but i can't find any method for detecting the collision between particular 2 boxes like "head and hand" or any other. 🙁
is there any method available?? :nerd:

Edit ::

Now i tried this but still not getting the result

if(MyInput.isPressed(MyInput.BUTTON1))
      {
         blue.setPos(blue.getX()+10, blue.getY());
//         orange.setFlipX(orange.getSkeleton().getFlipX()?false:true);
         blue.getbounds().update(blue.getSkeleton(),true);
         orange.getbounds().update(orange.getSkeleton(),true);
         if(blue.getbounds().aabbIntersectsSkeleton(orange.getbounds()))
         {
            Polygon p1 = new Polygon();
            Polygon p2 = new Polygon();
            Polygon p3 = new Polygon();
            for (int i = 0; i < blue.getbounds().getBoundingBoxes().size; i++){
               p1.setVertices(blue.getbounds().getBoundingBoxes().get(i).getVertices());
               for (int j = 0; j < orange.getbounds().getBoundingBoxes().size; j++){
                  p2.setVertices(orange.getbounds().getBoundingBoxes().get(j).getVertices());
                     if(Intersector .intersectPolygons(p1,p2,p3)){
                        System.out.println("blue : " +blue.getbounds().getBoundingBoxes().get(i).getName());
                        System.out.println("orange : " +orange.getbounds().getBoundingBoxes().get(ij.getName());
                     }
                  

           }
        }         
//            if (hit != null) {
//               System.out.println("hit: " + hit);
//               blue.getSkeleton().findSlot("head").getColor().set(Color.RED); // Turn head red until touchUp.
//            }
         }
         

  }

Thanks

Nate

SkeletonBounds collects BoundingBoxAttachments, transforms their vertices, and stores the resulting polygons. It has a few convenience methods for doing hit detection. If those don't do everything you need, you can access the polygons directly and do the hit detection yourself. If you keep your bounding boxes convex, you can use this code.

Note Street Fighter uses simpler bounding boxes:
https://www.google.hr/search?q=street+fighter+hit+boxes&tbm=isch
If simple rectangles are good enough for them it's probably fine for you. 🙂

ashwani

Hi nate,

Thanks it Worked 🙂

Westquote

Just thought I'd throw this out here in case people run into trouble with the limitation of Box2D's 8 vertices and/or convex hull. For our game, we use a library called 'poly2tri' (the one hosted on github) to perform a constrained Delaunay triangulation which very quickly outputs a variable number of triangles describing any polygon (which are guaranteed to be exactly 3 vertices, and thus be convex). With a caveats (such as language/environment limitations and needing to simplify certain degenerate polygons), it's worked perfectly for this and various other uses.

Also, to check overlap/distance/relationships of our hitboxes, we use b2QueryCallback, b2Distance, and b2RayCastCallback to determine various pieces of high-precision on-demand collision information.