the field concept and dependency graphs - uni-weimar.delopa1693/master/wise1516/field_concept... ·...

The Field Concept

and Dependency Graphs

Tim Weißker

Recap: Scene Graphs

hierarchical representation

of the elements of a scene

(and their properties) to

be rendered

2

simplified scene graph for a motorcycle

The Field Concept and Dependency Graphs 06.11.2015

Recap: Scene Graphs




be rendered

the field concept defines

every node of the scene

graph being a field container

3



Recap: Scene Graphs




be rendered

the field concept defines

every node of the scene

graph being a field container

But what does that mean?

4



Definitions

field

more complex form of an attribute

represents object state information

input and output fields

easy serialization and distribution

5 The Field Concept and Dependency Graphs 06.11.2015

Examples

6

avango avango.gua

SFBool SFMatrix4

SFInt SFVec3

SFFloat SFVec4

[…] […]

Fields


Definitions

7

field container

collection of fields

evaluate() method called when one field changes


Definitions

8

field container

collection of fields

evaluate() method called when one field changes

example:


Increment

Input : SFInt

Output : SFInt

evaluate()

class Increment(avango.script.Script):

Input = avango.SFInt()

Output = avango.SFInt()

def evaluate(self):

self.Output.value = self.Input.value + 1

Input : SFInt

Output : SFInt

evaluate()

def evaluate(self):


method evaluate()

a field container’s evaluate() method executes max. three steps

in the following order

Definitions


Input : SFInt

Output : SFInt

evaluate()

def evaluate(self):


method evaluate()



read values from local input fields

Definitions


Input : SFInt

Output : SFInt

evaluate()

def evaluate(self):


method evaluate()




calculate new values derived from the read values

Definitions


Input : SFInt

Output : SFInt

evaluate()

def evaluate(self):


method evaluate()





write the results to the output fields

Definitions


Input : SFInt

Output : SFInt

evaluate()

def evaluate(self):


method evaluate()





write the results to the output fields

to increase reuse, no external data should be accessed

Definitions


Examples

14

avango avango.gua

SFBool SFMatrix4

SFInt SFVec3

SFFloat SFVec4

[…] […]

avango.gua

TransformNode

GeometryNode

PointLightNode

Classes derived from avango.script.Script

Fields

Field Containers


Implementing a field container


class Container(avango.script.Script):




#declaration of fields, e.g.

sf_mat = avango.gua.SFMatrix4()






def __init__(self):

self.super(Container).__init__()






def __init__(self):


def my_constructor(PARAMETER1, PARAMETER2, …):

#initialize variables, parameters, etc.






def __init__(self):




def evaluate(self):

#perform update when fields change

Evaluation policies





def __init__(self):




def evaluate(self):

#perform update when fields change

Evaluation policies

21

evaluate()

called when at least one of the fields changes


Evaluation policies





def __init__(self):


self.always_evaluate(True)



def evaluate(self):

#perform update every frame

Evaluation policies

23

evaluate()



forces evaluation every frame regardless of field changes


Evaluation policies





def __init__(self):




@field_has_changed(sf_mat)

def sf_mat_changed(self):

#perform update when sf_mat changed

Evaluation policies

25

evaluate()



forces evaluation every frame regardless of field changes

@field_has_changed(SFFoo)

only evaluated when SFFoo changes

function name can vary


Dependencies

sometimes it is the case that an output field of one field

container forms the input of another one (e.g. sensors)


Dependencies



field connections: the value of a field is copied into

another one after evaluation


Dependency

Field connection

Node A

Input : type

Output : type

evaluate()

Node B

Input : type

Output : type

evaluate()

Dependencies



field connections: the value of a field is copied into

another one after evaluation

b.Input.connect_from(a.Output)


Dependency

Field connection

Node A

Input : type

Output : type

evaluate()

Node B

Input : type

Output : type

evaluate()

Dependencies

problem:

if node B is evaluated before

node A, the input of B is not

specified


Node A

Input : type

Output : type

evaluate()evaluateDependency()

Node B

Input : type

Output : type


Dependencies

problem:

if node B is evaluated before

node A, the input of B is not

specified

solution: evaluateDependency() method

is called on all unevaluated field containers

recursively calls same method on all dependent field containers

then calls evaluate() on itself

returns when recognizing a loop (warning message)


Node A

Input : type

Output : type


Node B

Input : type

Output : type


Feedback propagation

although evaluateDependency() checks and ignores cyclic

dependencies, they are sometimes unavoidable



although evaluateDependency() checks and ignores cyclic

dependencies, they are sometimes unavoidable

let‘s suppose we want to move a figure in a virtual

environment using an input device (e.g. joystick)




KeyboardInput

sf_move_left : SFBool

sf_move_right : SFBool

sf_jump : SFBool

sf_move_vec : SFVec3

evaluate()



KeyboardInput



sf_jump : SFBool


evaluate()

Accumulator


sf_mat : SFMatrix4

evaluate()


35

Combining the last

and relative

transformation

matrix in one


KeyboardInput



sf_jump : SFBool


evaluate()

Accumulator


sf_mat : SFMatrix4

evaluate()


36

Combining the last

and relative

transformation

matrix in one


KeyboardInput



sf_jump : SFBool


evaluate()

Accumulator


sf_mat : SFMatrix4

evaluate()

GroundFollowing

sf_mat : SFMatrix4

sf_corrected_mat :

SFMatrix4

evaluate()


37

Combining the last

and relative

transformation

matrix in one

Corrects the

transformation

matrix with respect

to collisions


KeyboardInput



sf_jump : SFBool


evaluate()

Accumulator


sf_mat : SFMatrix4

evaluate()

GroundFollowing

sf_mat : SFMatrix4

sf_corrected_mat :

SFMatrix4

evaluate()


38

Combining the last

and relative

transformation

matrix in one

Corrects the

transformation

matrix with respect

to collisions


KeyboardInput



sf_jump : SFBool


evaluate()

Accumulator


sf_mat : SFMatrix4

evaluate()

GroundFollowing

sf_mat : SFMatrix4

sf_corrected_mat :

SFMatrix4

evaluate()


39

Combining the last

and relative

transformation

matrix in one

Corrects the

transformation

matrix with respect

to collisions

problem: the output of GroundFollowing is required as an

input in the next frame to be rendered


KeyboardInput



sf_jump : SFBool


evaluate()

Accumulator


sf_mat : SFMatrix4

evaluate()

GroundFollowing

sf_mat : SFMatrix4

sf_corrected_mat :

SFMatrix4

evaluate()


40

solution: weak field connection


KeyboardInput



sf_jump : SFBool


evaluate()

Accumulator


sf_mat : SFMatrix4

evaluate()

GroundFollowing

sf_mat : SFMatrix4

sf_corrected_mat :

SFMatrix4

evaluate()


41

solution: weak field connection

is ignored during dependency evaluation

the corresponding value however is propagated in the next

frame accum.sf_mat.connect_weak_from(gf.sf_corrected_mat)


KeyboardInput



sf_jump : SFBool


evaluate()

Accumulator


sf_mat : SFMatrix4

evaluate()

GroundFollowing

sf_mat : SFMatrix4

sf_corrected_mat :

SFMatrix4

evaluate()

Transform nodes in the scenegraph


TransformNode

Name : SFString

Parent : SFNode

Children : MFNode

Transform : SFMatrix4

WorldTransform :

SFMatrix4

evaluate()

Node

Summary

43

every node of the scene graph and classes derived from avango.script.Script are called field container


Summary

44


state of a field container is defined by state of the fields

easy serialization

easy network distribution within AVANGO


Summary

45



easy serialization


within a field container, no external data is accessed

dependencies are realized with field connections

loose coupling between field container instances

easily reusable ( component-based design)


Summary

46



easy serialization


within a field container, no external data is accessed

dependencies are realized with field connections

loose coupling between field container instances

easily reusable ( component-based design)

feedback propagation using weak field connections


References

Kuck, R., Wind, J., Riege, K., Bogen, M.: Improving the

AVANGO VR/AR Framework – Lessons Learned, Paper from

the 5th GI VR/AR workshop, 2008

Fröhlich, B., Bernstein, A.C.: Avango –Virtual Reality

Framework, Virtual Reality lecture, 2011

Avango: http://www.avango.org/

Autodesk Maya Documentation: http://docs.autodesk.com

Guacamole simple example (VR Cluster):

/opt/avango/master/examples/simple_example


The End

48

Thank you for your attention!

In your next assignment, you will

build your first own dependency graph.

Have fun


the field concept and dependency graphs - uni-weimar.delopa1693/master/wise1516/field_concept... ·...

Documents