bRigid release - jbullet for processing

daniel.eds

bRigid release - jbullet for processing

in Library and Tool Development • 3 months ago

Hi,

I just released a library for an simpler handling of jbullet (rigid bodies) in processing:

bRigid provides classes for an easier handling of jBullet in Processing. bRigid is thought as a kind of Processing port for the bullet physics simulation library written in C++. This library allows the interaction of rigid bodies in 3D. Geometry/ Shapes are build with Processing PShape Class, for convinient display and export(dxf)

jBullet is a Java port of Bullet (c) 2008 Martin Dvorak http://jbullet.advel.cz/
Bullet Physics Library (c) 2003-2013 Erwin Coumans http://www.bulletphysics.com/
this work would be not possible without the jBullet/processing examples of Giulio Piacentino and Richard Brauer

I hope this could be an initial push, porting a physics library like bullet to processing.

you can download it:

for processing: http://www.lab-eds.org/bRigid
for coding: http://github.com/djrkohler/bRigid

Replies(12)

rbrauer

Re: bRigid release - jbullet for processing

3 months ago

This looks very impressive from the video. I'm excited to try it out soon. You might want to keep an eye on the JMonkey Engine project. I've looked into their Bullet code a little, but never used it. JBullet hasn't been updated for a while and the JMonkey version might have more features and less bugs. It looks like they are developing a Java interface on the cross platform native Bullet code as well and their modules are supposed to be usable apart from the rest of the engine.

daniel.eds

Re: Re: bRigid release - jbullet for processing

3 months ago

Yes, you are right, jBullet is a little bit outdated. But I could not figure out how you can use jME's-bullet.jar independently. It has a lot of dependencies to the other jME-packages. So I stayed with jBullet, as anyhow this lib is more a collection of sketches from the last 2-3 years. The release of p5 2.0 and its PShape class was the cause for putting everything into a library. It would be genius having a Java interface for native Bullet code. Especially for the upcoming bullet release, where all physics runs in openCl.

amnon.owed

Re: bRigid release - jbullet for processing

3 months ago

Awesome. Thanks for sharing!

>> Tweet

rbrauer

Re: bRigid release - jbullet for processing

3 months ago

Here is the code I used to get the bunny shown in my recent video in there. Thought it might be helpful. Its using Saito ObjLoader.

class BData {
  OBJModel bunny;
  
  BData() {
    bunny = new OBJModel(p5, "bunnyFlipD.obj");
    println(bunny.getFaceCount());
  }
  
  CollisionShape bshape() {
    int vertCount = bunny.getVertexCount();
    int faceCount = bunny.getFaceCount();
    ByteBuffer ver = ByteBuffer.allocateDirect(vertCount*12);
    ByteBuffer ind = ByteBuffer.allocateDirect(faceCount*12);
    for(int a=0; a<vertCount; a++) {
      PVector cur = bunny.getVertex(a);
//      ver.asFloatBuffer().put(cur.array());
      ver.putFloat(cur.x);
      ver.putFloat(cur.y);
      ver.putFloat(cur.z);
    }
    for(int a=0; a<faceCount; a++) {
      int[] cur = 
        bunny.getFaceInSegment(0, a).getVertexIndices();
//      ind.asIntBuffer().put(cur);
      ind.putInt(cur[0]);
      ind.putInt(cur[1]);
      ind.putInt(cur[2]);
    }
    TriangleIndexVertexArray tiva = 
      new TriangleIndexVertexArray(
        faceCount, ind, 12, vertCount, ver, 12);
    GImpactMeshShape gims = new GImpactMeshShape(tiva);
    gims.setLocalScaling(new Vector3f(40.0,40.0,40.0));
    gims.updateBound();
    GImpactCollisionAlgorithm.registerAlgorithm(
      (CollisionDispatcher)world.getDispatcher());
    return gims;
  }
  
  PShape getPShape() {
    PShape ret = createShape();
    ret.beginShape(TRIANGLES);
    for(int a=0; a<bunny.getFaceCount(); a++) {
      PVector[] cur = bunny.getFaceVertices(a);
      for(PVector vv : cur) {
        float rnd = random(212,255);
        ret.fill(rnd,255,rnd);
        vv.mult(40.0);
        ret.vertex(vv.x, vv.y, vv.z);
      }
    }
    ret.endShape();
    return ret;
  }
}

daniel.eds

Re: Re: bRigid release - jbullet for processing

3 months ago

nice example!

just tried to integrate it, but for some reason I get a "sticky" behavior.. uploaded it to github:

https://github.com/djrkohler/bRigid/commit/6cb87a15b45c3898add380cc1a57ccd7d9e6ca8f

rbrauer

Re: bRigid release - jbullet for processing

3 months ago

I think I know what you mean. Bullet seems to be very finicky about the magnitude of the numbers going into it as well as exactly how timesteps are set up. The sizes I'm using in that example are probably not ideal. Here are some good short articles with guidelines:

http://www.bulletphysics.org/mediawiki-1.5.8/index.php/Scaling_The_World
http://www.bulletphysics.org/mediawiki-1.5.8/index.php/Stepping_The_World

daniel.eds

Re: Re: bRigid release - jbullet for processing

3 months ago

mh.. don't have really success; scaled by factor .001-100. still looks like the bodies attract each other, not avoid collision. Having the right units was the trickiest part. I know the bodies are not in bullet sense in a right scale, they are too big. So everything is a little bit slow motion. But generally I'm more interested in packing and assembly and not collision itself. An over-scaled world helps me to keep the forces in some boudaries; damping and friction alone was here nether enough. Ah: one GImpactMesh works fine, just many of them freaks.. perhaps its not made for many?

rbrauer

Re: bRigid release - jbullet for processing

3 months ago

I have the feeling it works best with smaller amounts of triangles and triangles that don't have many smallish angles. It is probably quite computationally intense. You could also try making sure everything has friction set. I read somewhere that this makes it more stable and it doesn't give objects any friction by default.

Maybe this is the same issue, probably not fixed in the JBullet version:
https://code.google.com/p/bullet/issues/detail?id=662

daniel.eds

Re: Re: bRigid release - jbullet for processing

3 months ago

you are right. by adding explicitly friction to the rigidbodyObject the result becomes much more precise. thanks for the hint..

rbrauer

Re: bRigid release - jbullet for processing

3 months ago

I want to contribute some more example code. Here is what I used for a static triangle mesh. Some of this is maybe a little wonky, but basically you set up a TriangleIndexVertexArray again and give it to a BvhTriangleMeshShape collision shape. There is also code in here where I tried to animate the terrain like in the JBullet demo, but PShape setVertex(...) wasn't working when I tested it so its commented out.

class TData {
  PVector[] pvert;
  int count, countsq, countsqp, indCount, nsscroll;
  ByteBuffer ind, vert;
  PShape shape;
  
  TData(int count) {
    this.count = count;
    countsq = count*count;
    countsqp = countsq+(count*2+1);
    pvert = new PVector[countsqp];
    nsscroll = 1000;
  }
  
/*
  void update(float sz) {
    float hdim = sz*count*.5;
    float nscl = .09;
    float zscl = 7.5;
    for(int a=0; a<countsqp; a++) {
      int xi = a%(count+1);
      int yi = floor(a/(count+1));
      float xx = xi*sz-hdim;
      float zz = yi*sz-hdim;
      float yy = 0;
      yy += noise((nsscroll+xx+1)*nscl, zz*nscl);
      yy += noise((nsscroll+xx+1)*nscl, (zz+1)*nscl);
      yy += noise((nsscroll+xx-1)*nscl, zz*nscl);
      yy += noise((nsscroll+xx)*nscl, (zz-1)*nscl);
      yy *= .25*zscl;
//      float yy = noise(1000+xx*nscl, zz*nscl)*zscl;
      if(xi == 0 || xi == count || yi == 0 || yi == count)
        yy -= zscl*.2;
      pvert[a] = new PVector(xx, yy, zz);
//      vert.putFloat(a*4, xx);
//      vert.putFloat((a+1)*4, yy);
//      vert.putFloat((a+2)*4, zz);
      shape.setVertex(a, pvert[a]);
    }
    nsscroll++;
  }
//*/
  
  CollisionShape tshape(float sz) {
    float hdim = sz*count*.5;
    float nscl = .09;
    float zscl = 7.5;
    ind = ByteBuffer.allocateDirect(countsq*24);
    vert = ByteBuffer.allocateDirect(countsqp*12);
    for(int a=0; a<countsqp; a++) {
      int xi = a%(count+1);
      int yi = floor(a/(count+1));
      float xx = xi*sz-hdim;
      float zz = yi*sz-hdim;
      float yy = 0;
      yy += noise((nsscroll+xx+1)*nscl, zz*nscl);
      yy += noise((nsscroll+xx+1)*nscl, (zz+1)*nscl);
      yy += noise((nsscroll+xx-1)*nscl, zz*nscl);
      yy += noise((nsscroll+xx)*nscl, (zz-1)*nscl);
      yy *= .25*zscl;
//      float yy = noise(1000+xx*nscl, zz*nscl)*zscl;
      if(xi == 0 || xi == count || yi == 0 || yi == count)
        yy -= zscl*.2;
      pvert[a] = new PVector(xx, yy, zz);
      vert.putFloat(xx);
      vert.putFloat(yy);
      vert.putFloat(zz);
      if(a != 0 && (xi == count || a > countsq+(count-2))) {
        continue;
      }
      indCount += 2;
      ind.putInt(a);
      ind.putInt(a+1);
      ind.putInt(a+count+1);
      ind.putInt(a+1);
      ind.putInt(a+count+2);
      ind.putInt(a+count+1);
    }
    TriangleIndexVertexArray tiva = 
      new TriangleIndexVertexArray(
        indCount, ind, 12, countsq, vert, 12);
        
    return new BvhTriangleMeshShape(tiva, true);
  }
  
  PShape getPShape() {
    if(shape != null) return shape;
    ByteBuffer indCopy = ind.duplicate();
    indCopy.rewind();
    PShape ret = createShape();
    ret.beginShape(TRIANGLES);
//    ret.stroke(80);
    for(int a=0; a<countsq*6; a++) {
      float rnd = random(168,255);
      int cind = indCopy.getInt();
      PVector cur = pvert[cind];
      ret.fill(rnd-24,rnd,224);
      ret.vertex(cur.x, cur.y, cur.z);
    }
    ret.endShape();
    shape = ret;
    return ret;
  }

daniel.eds

Re: Re: bRigid release - jbullet for processing

3 months ago

wow, looks promising. I always was pondering over the TriangleIndexVertexArray.. now I've got it..

I will implement it soon, but approx. just in two weeks.

thanks for sharing. great! Or generally I'm interested in your opinion what could be the net step to build up a serious physics library based on bullet?

rbrauer

Re: bRigid release - jbullet for processing

3 months ago

I haven't even tried all the JBullet features you have implemented like joints, constraints, and motors and haven't used your library yet. I'm browsing the javadoc now. I've been fiddling with other projects, but I definitely want to get into this and I might have some thoughts a little later.

Top Reply