[code] Verlet Rope « cocos2d for iPhone

cocos2d for iPhone » Programming » physics

[code] Verlet Rope

(69 posts) (26 voices)

Started 1 year ago by patrickC
Latest reply from flightless

Tags:

12 3 Next »

patrickC
Member

Hey all,
just thought I'd share a small set of classes to draw a Verlet rope that people might find useful.

I created it to use in conjunction with box2d's new b2RopeJoint (which constricts two bodies with a "rope" joint, but obviously has no visual representation), although you can easily use the rope in other scenarios, as long as you feed it a start and end end point at creation and update.

By using VRope you can easily draw a segmented rope updated with a fast Verlet integration

Be aware that the "draw" method simply uses cocos2d's ccDrawLine method, other people might want to find more elegant solutions.
As always, my code is far from optimized but seems to suit my needs, so use at own risk :)

Download: http://cleverhamstergames.com/VRope-0.1.zip

USAGE
#import "VRope.h"

pointA and pointB are start and end points of the rope

Create
VRope *verletRope = [[VRope alloc] init:pointA pointB:pointB];
Update
[verletRope updateRope:dt pointA:pointA pointB:pointB];
Draw
[verletRope draw];

Posted 1 year ago #
GhostDog
Member

Thanks for this. Would fit nicely in game I'm working on.

Posted 1 year ago #
mobilebros
Moderator

Very cool @patrickC, i'll try it out as soon as i can!

Posted 1 year ago #
cocos
Member

Thank you @patrickC for sharing! That tells a lot about you!

Posted 1 year ago #
patrickC
Member

Small update to the verlet rope class.

Usage "instructions" are in the header file of the main class.
http://cleverhamstergames.com/VRope-0.2.zip

As always, use at own risk, etc...

CHANGES IN 0.2:
- Added CCSPriteSheet to display rope segments with a sprite (using GL_REPEAT)
- Added an anti-sagness "hack"
- Minor bug fix

Posted 1 year ago #
gavinb
Member

wow, great work @patrickC, much appreciated

Posted 1 year ago #
mobilebros
Moderator

Hey @patrickC, I tried this out with chipmunk's cpSlideJoint and it's looks great! I noticed there may be some small memory leaks particularly with allocating VStick and VPoint and having no corresponding release/autorelease (if I remember correctly)...

Posted 1 year ago #
Ozden79
Member

Hi There,

The link for v0.2 seems to be broken or is it only me?

Özden

Posted 1 year ago #
patrickC
Member

@mobilebros thanks for pointing that out, should be fixed now.

To avoid confusion, the latest version of VRope from now on will be located at:
http://cleverhamstergames.com/dev/vrope-latest.zip

Changes in 0.3:
- The class is now designed to be used with box2d bodies, but also has helper method to use it without box2d
- memory leaks fixed
- minor improvements

Also, check out the article I wrote on using VRope
http://www.cocos2d-iphone.org/archives/1112

Posted 1 year ago #

indy2005
Member

Hi,

Thanks for this. I am trying to replace Box2D into LevelSVG as you suggested - is it as simple as swapping one for the other, as I had always assumed perhaps that Riq did something to the implementation to make it Cocos2D compatible (I am likely very wrong!).

I just integrated Box2D as you suggest into my existing LevelSVG game and have the following errors:

'class b2PolygonShape' has no member named 'SetAsEdge' when using the method:

pathShape.SetAsEdge(currentPoint, bezierVerts[i]);

In b2World DrawShape:

void b2World::DrawShape(b2Fixture* fixture, const b2Transform& xf, const b2Color& color)
{
	switch (fixture->GetType())  //  warning: enumeration value 'e_unknown' not handled in switch

and in b2ContactSolver (see last line below)

void b2ContactSolver::InitializeVelocityConstraints()
{
	for (int32 i = 0; i < m_count; ++i)
	{
		b2ContactConstraint* cc = m_constraints + i;

		float32 radiusA = cc->radiusA;
		float32 radiusB = cc->radiusB;
		b2Body* bodyA = cc->bodyA;
		b2Body* bodyB = cc->bodyB;
		b2Manifold* manifold = cc->manifold;

		b2Vec2 vA = bodyA->m_linearVelocity;
		b2Vec2 vB = bodyB->m_linearVelocity;
		float32 wA = bodyA->m_angularVelocity;
		float32 wB = bodyB->m_angularVelocity;

		b2Assert(manifold->pointCount > 0);

		b2WorldManifold worldManifold;
		worldManifold.Initialize(manifold, bodyA->m_xf, radiusA, bodyB->m_xf, radiusB);

		cc->normal = worldManifold.normal;

		for (int32 j = 0; j < cc->pointCount; ++j)
		{
			b2ManifoldPoint* cp = manifold->points + j;   // WARNING: unused variable 'cp'

Any advice appreciated

Regards

Posted 1 year ago #

patrickC
Member

The error you're getting is because the new version of box2d has some api changes and edges are created in a different way. (but I haven't looked into that yet)
the other two are warnings that don't influence functionality.
updating box2d is as easy as swapping one for the other, obviously taking into account any api changes ;)

Posted 1 year ago #
indy2005
Member

indeed....there is a new b2EdgeShape used to created edges rather than using b2PolygonShape.SetAsEdge.....

The World::Draw error seems to be because there are 2 new shapes, b2EdgeShape and b2LoopShape, and the switch statement doesnt have a default statement, and doesnt deal with these 2 new shapes...

I didn't know switch statements were that clever?? i.e. to know when you arent checking every value in an enum?

Regards

i

Posted 1 year ago #
joep
Member

Has anyone integrated Box2d and Chipmunk in same Cocos project ?
eg. dropping this code into existing cocos/chipmunk project.

Thanks !

Posted 1 year ago #
patrickC
Member

@joep
You don't need to use box2d with VRope.
To use it with chipmunk (or anything else also) remove the Box2D.h import from VRope.h and use the initWithPoints and updateWithPoints methods instead of the regular init and update methods (that take box2d bodies as arguments).
Using the "withPoints" methods you need to pass the coordinates of the start and end point of the rope.

Posted 1 year ago #
charlie_fulton
Member

@patrickC thank you so much for this contribution! Your article is great.

Posted 1 year ago #
indy2005
Member

Hi,

Whats the general technique for deleting the rope joint and verlet visual representation?

Regards

i

Posted 1 year ago #

patrickC
Member

From the VRope class header:

REMOVING:
To remove a rope you need to call the removeSprites method and then release:
[verletRope removeSprites]; //remove the sprites of this rope from the spritebatchnode
[verletRope release];

VRope is not responsible for creating or destroying the actual box2d joint.
To destroy a box2d joint use DestroyJoint:

world->DestroyJoint(joint);

Hope that helps,
cheers

Posted 1 year ago #

indy2005
Member

thanks!

Posted 1 year ago #
obscenum
Member

Good work, thanks.

Posted 1 year ago #
indy2005
Member

Hi,

Can I ask, if I wanted the rope to attach to the top of a Box2D body, not the centre - regardless of its orientation (i.e. the b2Bodies local space top), can you suggest the best approach?

Regards

i

Posted 1 year ago #
indy2005
Member

bump ;-)

...using the localanchor points doesn't seem to work very well for me.

Posted 1 year ago #
patrickC
Member
@indy2005
VRope and the b2RopeJoint are two "separate" things.
If you want to attach a b2RopeJoint at the top of a body, set the anchor points appropriately when creating the joint, which will make sure box2d simulates the joint correctly.
You'll then need to make some modifications to VRope so that it takes into account the local anchor point of the joint, rather than using the body's center as it is now.

I would change the init function to also take an extra parameter, the b2RopeJoint that you've just created (not the b2RopeJointDef), and instead of using body1 and body2's world center, use the two local anchor points of the joint in world coordinates
```
-(id)init:(b2Body*)body1 body2:(b2Body*)body2 spriteSheet:(CCSpriteBatchNode*)spriteSheetArg ropeJoint:(b2RopeJoint*)ropeJoint {
   ...
   ...
   CGPoint pointA = ccp(ropeJoint->GetAnchorA().x*PTM_RATIO,ropeJoint->GetAnchorA().y*PTM_RATIO);
   CGPoint pointB = ccp(ropeJoint->GetAnchorB().x*PTM_RATIO,ropeJoint->GetAnchorB().y*PTM_RATIO);
   ...
```
You will also need to update the resetRope method, if you're using it, to replicate the behavior (resetRope also uses body's center at the moment).

If the GetAnchor methods of the box2d joint are returning the expected result (anchor points in world coordinates) it should work, but as you might guess I just wrote this up now and haven't tested it.
Let me know if you're having problems,
cheers
Posted 1 year ago #
indy2005
Member

Hi,

Thanks will try when I get back, had already tried a new init() method with an offset parameter with no success.

Do you know if the local anchor (for example (0,1) is always the same position in the body regardless of the bodies rotation? i.e. if a body is on its side, (0,1) always represents the same "real" point on the body i.e the local co-ordinate system which rotates around with the body??

Local and World points always confuse me...

Regards

i

Posted 1 year ago #
indy2005
Member

Hi,

Seems to work a treat, thanks...although you have to update reset() and update() in the same way, and keep a reference to the rope joint as an instance variable so it can be reused outside of the init().

Would still like to know how local co-ordinates work ;-)

Regards

i

Posted 1 year ago #
indy2005
Member

Hi,

me again ;-)

Are you getting any odd behaviour i you let an object freefall and then use the rope? My verlet rope (when using long distances), doesnt mirror the constraint being applied by the rope. i.e. I have a lot of additional slack in my rope once the max distance of the joint is reached?

Regards

i

Posted 1 year ago #
patrickC
Member

@indy2005
glad to hear you got something working.
Your reasoning on the local anchor point seems to be correct (I think...), say you have a point at the top of a box [0,1], rotate it 180 degrees, that point will now be visually at the bottom of the box but, in local coordinates, it's still at [0,1].

The more segments you have in a rope and the more sag/slack will be introduced.
You need to play with segment length, number of iterations when solving the points, and also the magic number of the "anti sag" hack.
Unfortunately this type of verlet point solving introduces a certain amount of slack that can't be avoided, unless you solve all the points at once, beyond my comprehension :)

Posted 1 year ago #
mobilebros
Moderator

@indy2005, Regarding the slack - I noticed the same thing and after playing with the 0.9 value with the "NOTICE" above it (in the init and reset methods) and also the number of iterations of the stick contractions (update method), I was able to get "better" results.

I think I used 0.85 and 6 respectively.

Posted 1 year ago #
indy2005
Member

Hi thanks all.

@mobilebros, I can see the sticks variable, but cant see anything in the init or reset of VRope?

Regards

i

Posted 1 year ago #

delimy
Member

@patrickC, thank you for your share!
here's a few questions:

[[vPoints objectAtIndex:i] applyGravity:dt];
		[[vPoints objectAtIndex:i] update];

y was changed in applyGravity, so the oldy in update is not really the "old" y.
from what I learned here,
y = y + y -oldy + (-10) * dt *dt;

but you are using (-10) * dt in applyGravity.

thus I combine the two methods into one like this:

float tempx = x;
	float tempy = y;
	x += x - oldx;
	y += y - oldy -10.0f*dt*dt;
	oldx = tempx;
	oldy = tempy;

There's no visual-able difference in this test project, however.

Another issue I found, was that there's no interaction between the two systems.
e.g. bodyB will not be "pulled up" by an over-tight rope.

After some tests, I got this:

-(void)updateWithPoints:(CGPoint)pointA pointB:(CGPoint)pointB dt:(float)dt {
	//manually set position for first and last point of rope
	[[vPoints objectAtIndex:0] setPos:pointA.x y:pointA.y];
	[[vPoints objectAtIndex:numPoints-1] setPos:pointB.x y:pointB.y];

	//update points, apply gravity
	for(int i=1;i<numPoints;i++) { // count on the last point here
		[[vPoints objectAtIndex:i] applyGravity:dt];
		[[vPoints objectAtIndex:i] update];
	}

	//contract sticks
	int iterations = 4;
	for(int j=0;j<iterations;j++) {
		for(int i=0;i<numPoints-1;i++) {
			[[vSticks objectAtIndex:i] contract];
		}
	}
  VPoint *last = [vPoints objectAtIndex:numPoints-1];
  b2Vec2 nb(last.x/32, last.y/32);
  b2Vec2 diff = nb - bodyB->GetPosition();
  diff = bodyB->GetMass() * 2 * diff; // larger value make the rope more elastic
  bodyB->ApplyLinearImpulse(diff, bodyB->GetPosition());
}

My question is how to calculate the correct impulse on bodyB (mine just happens to look OK but really make nonsense in physics).

Posted 1 year ago #

rohn
Member

Hi @patrickC - thanks for sharing the Verlet Rope, it's a great addition!

My question is about interacting with the rope. I read it can't collide with objects, but is it possible to know if I "touched" it?
I was trying to implement an approach using CGRectContainsPoint but was having trouble with where/how to define it using CGRectMake. I assumed I'd do that right before creating a new rope, after creating the b2RopeJoint.

Or is there a better way you'd suggest?

Edit (for clarification): the rectangle approach didn't seem right because I realized that the rectangle changes width (and height) as the rope/object goes through its swing.

Thanks!

Posted 1 year ago #

RSS feed for this topic

12 3 Next »

Reply »

You must log in to post.