| 1 | ||
|
Editor: antont
Time: 2006/05/29 20:59:46 GMT+0 |
||
| Note: | ||
changed: - a little test with an early blender tool by macouno, http://studio.kyperjokki.fi/data/swarm/0001_0101.avi blend for that stupidity is <a href="swarm3.blend">swarm3.blend</a> <PRE> @path "classes" @include "Stationary.tz" @include "Control.tz" @include "Mobile.tz" @define SWARM_SIZE 80. # # This simulation is inspired by Craig Reynolds classic Boids flocking # algorithm. The implementation is a bit more complicated than some of # the other breve demos. # # The idea behind the algorithm is that each bird is influenced by three # main urges: # 1) the urge to avoid collisions with neighbors # 2) the urge to stay close to neighbors # 3) the urge to match the velocity of neighbors # # These three urges are implemented in the Birds class, and by giving # them different weights, we produce different flocking behaviors. # Additionally, this simulation adds a "random" urge, and an urge # to stay close to the center of the world. # # If a bird touches the ground, it will land and stay on the ground # for a short time before rejoining the flock. # # Run the demo as it is and try the different behaviors from the # Simulation menu. Then try modifying the urge weights in the # Birds class yourself to change the behavior of the flock. # # This demo was O(n^2) previous to breve 1.2. We used to make every # bird look at the location of every other bird in order to assess # whether it was close enough to have an influence on behavior. Starting # in breve version 1.2 we support "neighborhoods", a very fast and # efficient way for objects in the simulation to find other objects # within a certain range. # # It is unfortunate that sometimes it is difficult to see all the birds at # once, and that if the flock breaks in two, the camera may point in between # them such that neither is visible. # Controller Swarm. Control : Swarm { + variables: birds (list). item (object). breveTexture (int). cloudTexture (int). selection (object). # the menu items corresponding to different flocking behaviors wackyMenu (object). obedientMenu (object). normalMenu (object). + to click on item (object): if selection: selection hide-neighbor-lines. if item: item show-neighbor-lines. selection = item. super click on item. + to init: floor (object). # Set up menus to modify the type of movement allowed. self add-menu named "Smoosh The Birdies" for-method "squish". self add-menu-separator. obedientMenu = (self add-menu named "Flock Obediently" for-method "flock-obediently"). normalMenu = (self add-menu named "Flock Normally" for-method "flock-normally"). wackyMenu = (self add-menu named "Flock Wackily" for-method "flock-wackily"). self enable-lighting. self enable-smooth-drawing. self move-light to (0, 20, 20). breveTexture = (self load-image from "breve.sgi"). cloudTexture = (self load-image from "clouds.sgi"). # Add a huge floor. floor = new Stationary. floor register with-shape (new Shape init-with-cube size (100, 2, 100)) at-location (0, -5, 0). floor catch-shadows. birds = SWARM_SIZE new Birds. birds set-bitmap to breveTexture. # we'll call this method to set up the initial bird behavior. self flock-normally. self set-background-texture to cloudTexture. self offset-camera by (5, 1.5, 6). self enable-shadows. + to iterate: location (vector). topDiff (double). highLoc (vector). self update-neighbors. # get the average location, and point the camera at it. foreach item in birds: { item fly. location += (item get-location). if | (item get-location) | > | highLoc |: highLoc = (item get-location). } # in the context of a number, the list represents the number of items # in the list. location /= birds. # print "the highest location is $location". topDiff = 0.0. foreach item in birds: { if topDiff < |location - (item get-location) |: topDiff = | location - (item get-location) |. } # we'll try to keep the camera aimed at the center of the # flock, and zoom in and out depending on how spread out the # flock is. it's doesn't work very well :). self aim-camera at location. self zoom-camera to (.5 * topDiff) + 10. # call the superclass iterate method to step the simulation forward. super iterate. # The following methods are the menu items we provide. + to squish: birds move to (0, 0, 0). + to flock-normally: birds flock-normally. normalMenu check. obedientMenu uncheck. wackyMenu uncheck. + to flock-obediently: birds flock-obediently. normalMenu uncheck. obedientMenu check. wackyMenu uncheck. + to flock-wackily: birds flock-wackily. normalMenu uncheck. obedientMenu uncheck. wackyMenu check. } Mobile : Birds { + variables: landed (int). # These are all parameters that can change the way the flock # behaves. Look at the flock-normally method for an example. cruiseDistance (float). maxAcceleration (float). maxVelocity (float). wanderConstant (float). worldCenterConstant (float). centerConstant (float). velocityConstant (float). spacingConstant (float). + to init: # register the object, set the initial location, velocity and color. self register with-shape ((new Shape) init-with-sphere radius .1). self move to random[(10, 10, 10)] - (5, -5, 5). self set-velocity to random[(20, 20, 20)] - (10, 10, 10). self set-color to random[(1, 1, 1)]. # when we hit the ground, we want to land on it using the method "land". # it gets called automatically when the collision occurs. self handle-collisions with-type "Stationary" with-method "land". # set the neighborhood radius that we will look in to find neighbors. self set-neighborhood-size to 3.0. + to flock-normally: wanderConstant = 4.0. worldCenterConstant = 5.0. centerConstant = 2.0. velocityConstant = 2.0. spacingConstant = 5.0. maxVelocity = 15. maxAcceleration = 15. cruiseDistance = .4. + to flock-normally-old: wanderConstant = 6.0. worldCenterConstant = 5.0. centerConstant = 2.0. velocityConstant = 4.0. spacingConstant = 5.0. maxVelocity = 20. maxAcceleration = 12. cruiseDistance = .4. + to flock-obediently: # The obedient flocking has a lower maxAcceleration, which means that # the objects don't react as quickly. This has a calming effect on the # agents. wanderConstant = 6.0. worldCenterConstant = 6.0. centerConstant = 2.0. velocityConstant = 3.0. spacingConstant = 4.0. maxVelocity = 16. maxAcceleration = 20. cruiseDistance = 1. + to flock-wackily: # The wacky flocking has a high maximum acceleration so that the # agents act heavily on all of their impulses every iteration. # This makes them swarm like insects. wanderConstant = 8.0. worldCenterConstant = 14.0. centerConstant = 1.0. velocityConstant = 3.0. spacingConstant = 4.0. maxVelocity = 20. maxAcceleration = 30. cruiseDistance = .5. + to land with ground (object): # if we hit the ground, we stop moving and set the landed flag to 1. # print "Landing". self set-acceleration to (0, 0, 0). self set-velocity to (0, 0, 0). landed = 1. # we don't want to keep colliding with the ground, or else we'll # keep on getting stuck again at every iteration--move up just a # tiny bit. self offset by (0, 0.1, 0). + to check-landed: # Ask the object if it has landed... return landed. + to fly: # This is the method which does it all. bird (object). toNeighbor (vector). centerUrge (vector). worldCenterUrge (vector). velocityUrge (vector). spacingUrge (vector). wanderUrge (vector). acceleration (vector). newVelocity (vector). neighbors (list). take-off (int). # get a list of neighbors within our neighborhood-size # (set during init), and check to see if they're visible. # using get neighbors, we very quickly get a list of # eligible objects and then check them further. this is # *much* faster than checking every bird in the simulation. foreach bird in (self get-neighbors): if (self check-visibility of bird): push bird onto neighbors. # if we're on the ground now, we'll pick a random number to see if we # take off again. The chances are 1 in 40, which is actually pretty # reasonable since this code is called every iteration. if landed: { take-off = random[40]. if take-off == 1: { # if we decide to take off, pick a random direction, # but not towards the ground. landed = 0. self set-velocity to random[(.1, 1.1, .1)] - (.05, 0, .05). } else { return. } } # get the urge towards the center, and velocity matching urge. centerUrge = (self get-center-urge with neighbors). velocityUrge = (self get-velocity-urge with neighbors). # are we too close to our neighbors? get a vector which reflects that # urge. foreach bird in neighbors: { toNeighbor = (self get-location) - (bird get-location). if |toNeighbor| < cruiseDistance: spacingUrge += toNeighbor. } # Are we wandering too far from the center of the world? if |(self get-location)| > 10: worldCenterUrge = -(self get-location). # Add a random component to the accumulation. wanderUrge = random[(2, 2, 2)] - (1, 1, 1). # normalize all of the vectors to length 1. if |spacingUrge|: spacingUrge /= |spacingUrge|. if |worldCenterUrge|: worldCenterUrge /= |worldCenterUrge|. if |velocityUrge|: velocityUrge /= |velocityUrge|. if |centerUrge|: centerUrge /= |centerUrge|. if |wanderUrge|: wanderUrge /= |wanderUrge|. wanderUrge *= wanderConstant. worldCenterUrge *= worldCenterConstant. centerUrge *= centerConstant. velocityUrge *= velocityConstant. spacingUrge *= spacingConstant. acceleration = (worldCenterUrge + centerUrge + velocityUrge + spacingUrge + wanderUrge). if |acceleration| != 0: acceleration /= |acceleration|. self set-acceleration to maxAcceleration * acceleration. newVelocity = (self get-velocity). if |newVelocity| > maxVelocity: newVelocity = maxVelocity * newVelocity/|newVelocity|. self set-velocity to newVelocity. + to get-velocity-urge with flock (list): item (object). count (float). velocity (vector). # get the average velocity of all the visible birds in the flock. foreach item in flock: { count += 1. velocity += (item get-velocity). } if count == 0: return (0, 0, 0). velocity /= count. return velocity - (self get-velocity). + to get-center-urge with flock (list): item (object). count (float). center (vector). # get the average location of all the visible birds in the flock. foreach item in flock: { count += 1. center += (item get-location). } if count == 0: return (0, 0, 0). center /= count. return center - (self get-location). + to check-visibility of item (object): # An item is visible if it is within a certain angle (2.0 radians) # of the direction we're facing (assumed to be a vector in the # direction we're moving). if (item == self): return 0. if !(item is a "Birds"): return 0. if (self get-angle to item) > 2.0: return 0. if (item check-landed): return 0. return 1. + to get-angle to otherMobile (object): tempVector (vector). tempVector = (otherMobile get-location) - (self get-location). return angle((self get-velocity), tempVector). } </PRE>
a little test with an early blender tool by macouno, http://studio.kyperjokki.fi/data/swarm/0001_0101.avi
blend for that stupidity is swarm3.blend
@path "classes"@include "Stationary.tz" @include "Control.tz" @include "Mobile.tz"
@define SWARM_SIZE 80.
# # This simulation is inspired by Craig Reynolds classic Boids flocking # algorithm. The implementation is a bit more complicated than some of # the other breve demos. # # The idea behind the algorithm is that each bird is influenced by three # main urges: # 1) the urge to avoid collisions with neighbors # 2) the urge to stay close to neighbors # 3) the urge to match the velocity of neighbors # # These three urges are implemented in the Birds class, and by giving # them different weights, we produce different flocking behaviors. # Additionally, this simulation adds a "random" urge, and an urge # to stay close to the center of the world. # # If a bird touches the ground, it will land and stay on the ground # for a short time before rejoining the flock. # # Run the demo as it is and try the different behaviors from the # Simulation menu. Then try modifying the urge weights in the # Birds class yourself to change the behavior of the flock. # # This demo was O(n^2) previous to breve 1.2. We used to make every # bird look at the location of every other bird in order to assess # whether it was close enough to have an influence on behavior. Starting # in breve version 1.2 we support "neighborhoods", a very fast and # efficient way for objects in the simulation to find other objects # within a certain range. # # It is unfortunate that sometimes it is difficult to see all the birds at # once, and that if the flock breaks in two, the camera may point in between # them such that neither is visible. #
Controller Swarm.
Control : Swarm { + variables: birds (list). item (object). breveTexture (int). cloudTexture (int). selection (object).
# the menu items corresponding to different flocking behaviors
wackyMenu (object). obedientMenu (object). normalMenu (object).
+ to click on item (object): if selection: selection hide-neighbor-lines. if item: item show-neighbor-lines.
selection = item.
super click on item.
+ to init: floor (object).
# Set up menus to modify the type of movement allowed.
self add-menu named "Smoosh The Birdies" for-method "squish". self add-menu-separator. obedientMenu = (self add-menu named "Flock Obediently" for-method "flock-obediently"). normalMenu = (self add-menu named "Flock Normally" for-method "flock-normally"). wackyMenu = (self add-menu named "Flock Wackily" for-method "flock-wackily").
self enable-lighting. self enable-smooth-drawing. self move-light to (0, 20, 20).
breveTexture = (self load-image from "breve.sgi"). cloudTexture = (self load-image from "clouds.sgi").
# Add a huge floor.
floor = new Stationary. floor register with-shape (new Shape init-with-cube size (100, 2, 100)) at-location (0, -5, 0). floor catch-shadows.
birds = SWARM_SIZE new Birds.
birds set-bitmap to breveTexture.
# we'll call this method to set up the initial bird behavior.
self flock-normally.
self set-background-texture to cloudTexture.
self offset-camera by (5, 1.5, 6).
self enable-shadows.
+ to iterate: location (vector). topDiff (double). highLoc (vector).
self update-neighbors.
# get the average location, and point the camera at it.
foreach item in birds: { item fly. location += (item get-location).
if | (item get-location) | > | highLoc |: highLoc = (item get-location). }
# in the context of a number, the list represents the number of items # in the list.
location /= birds. # print "the highest location is $location".
topDiff = 0.0.
foreach item in birds: { if topDiff < |location - (item get-location) |: topDiff = | location - (item get-location) |. }
# we'll try to keep the camera aimed at the center of the # flock, and zoom in and out depending on how spread out the # flock is. it's doesn't work very well :).
self aim-camera at location. self zoom-camera to (.5 * topDiff) + 10.
# call the superclass iterate method to step the simulation forward.
super iterate.
# The following methods are the menu items we provide.
+ to squish: birds move to (0, 0, 0).
+ to flock-normally: birds flock-normally. normalMenu check. obedientMenu uncheck. wackyMenu uncheck.
+ to flock-obediently: birds flock-obediently. normalMenu uncheck. obedientMenu check. wackyMenu uncheck.
+ to flock-wackily: birds flock-wackily. normalMenu uncheck. obedientMenu uncheck. wackyMenu check. }
Mobile : Birds { + variables: landed (int).
# These are all parameters that can change the way the flock # behaves. Look at the flock-normally method for an example.
cruiseDistance (float).
maxAcceleration (float). maxVelocity (float).
wanderConstant (float). worldCenterConstant (float). centerConstant (float). velocityConstant (float). spacingConstant (float).
+ to init: # register the object, set the initial location, velocity and color.
self register with-shape ((new Shape) init-with-sphere radius .1). self move to random[(10, 10, 10)] - (5, -5, 5). self set-velocity to random[(20, 20, 20)] - (10, 10, 10). self set-color to random[(1, 1, 1)].
# when we hit the ground, we want to land on it using the method "land". # it gets called automatically when the collision occurs.
self handle-collisions with-type "Stationary" with-method "land".
# set the neighborhood radius that we will look in to find neighbors.
self set-neighborhood-size to 3.0.
+ to flock-normally: wanderConstant = 4.0. worldCenterConstant = 5.0. centerConstant = 2.0. velocityConstant = 2.0. spacingConstant = 5.0.
maxVelocity = 15. maxAcceleration = 15. cruiseDistance = .4.
+ to flock-normally-old: wanderConstant = 6.0. worldCenterConstant = 5.0. centerConstant = 2.0. velocityConstant = 4.0. spacingConstant = 5.0.
maxVelocity = 20. maxAcceleration = 12. cruiseDistance = .4.
+ to flock-obediently: # The obedient flocking has a lower maxAcceleration, which means that # the objects don't react as quickly. This has a calming effect on the # agents.
wanderConstant = 6.0. worldCenterConstant = 6.0. centerConstant = 2.0. velocityConstant = 3.0. spacingConstant = 4.0.
maxVelocity = 16. maxAcceleration = 20. cruiseDistance = 1.
+ to flock-wackily: # The wacky flocking has a high maximum acceleration so that the # agents act heavily on all of their impulses every iteration. # This makes them swarm like insects.
wanderConstant = 8.0. worldCenterConstant = 14.0. centerConstant = 1.0. velocityConstant = 3.0. spacingConstant = 4.0.
maxVelocity = 20. maxAcceleration = 30. cruiseDistance = .5.
+ to land with ground (object): # if we hit the ground, we stop moving and set the landed flag to 1.
# print "Landing".
self set-acceleration to (0, 0, 0). self set-velocity to (0, 0, 0).
landed = 1.
# we don't want to keep colliding with the ground, or else we'll # keep on getting stuck again at every iteration--move up just a # tiny bit.
self offset by (0, 0.1, 0).
+ to check-landed: # Ask the object if it has landed...
return landed.
+ to fly: # This is the method which does it all.
bird (object). toNeighbor (vector).
centerUrge (vector). worldCenterUrge (vector). velocityUrge (vector). spacingUrge (vector). wanderUrge (vector).
acceleration (vector).
newVelocity (vector).
neighbors (list).
take-off (int).
# get a list of neighbors within our neighborhood-size # (set during init), and check to see if they're visible. # using get neighbors, we very quickly get a list of # eligible objects and then check them further. this is # much faster than checking every bird in the simulation.
foreach bird in (self get-neighbors): if (self check-visibility of bird): push bird onto neighbors.
# if we're on the ground now, we'll pick a random number to see if we # take off again. The chances are 1 in 40, which is actually pretty # reasonable since this code is called every iteration.
if landed: { take-off = random[40].
if take-off == 1: { # if we decide to take off, pick a random direction, # but not towards the ground.
landed = 0. self set-velocity to random[(.1, 1.1, .1)] - (.05, 0, .05). } else { return. } }
# get the urge towards the center, and velocity matching urge.
centerUrge = (self get-center-urge with neighbors). velocityUrge = (self get-velocity-urge with neighbors).
# are we too close to our neighbors? get a vector which reflects that # urge.
foreach bird in neighbors: { toNeighbor = (self get-location) - (bird get-location). if |toNeighbor| < cruiseDistance: spacingUrge += toNeighbor. }
# Are we wandering too far from the center of the world?
if |(self get-location)| > 10: worldCenterUrge = -(self get-location).
# Add a random component to the accumulation.
wanderUrge = random[(2, 2, 2)] - (1, 1, 1).
# normalize all of the vectors to length 1.
if |spacingUrge|: spacingUrge /= |spacingUrge|. if |worldCenterUrge|: worldCenterUrge /= |worldCenterUrge|. if |velocityUrge|: velocityUrge /= |velocityUrge|. if |centerUrge|: centerUrge /= |centerUrge|. if |wanderUrge|: wanderUrge /= |wanderUrge|.
wanderUrge = wanderConstant. worldCenterUrge = worldCenterConstant. centerUrge = centerConstant. velocityUrge = velocityConstant. spacingUrge *= spacingConstant.
acceleration = (worldCenterUrge + centerUrge + velocityUrge + spacingUrge + wanderUrge).
if |acceleration| != 0: acceleration /= |acceleration|. self set-acceleration to maxAcceleration * acceleration.
newVelocity = (self get-velocity).
if |newVelocity| > maxVelocity: newVelocity = maxVelocity * newVelocity/|newVelocity|.
self set-velocity to newVelocity.
+ to get-velocity-urge with flock (list): item (object). count (float). velocity (vector).
# get the average velocity of all the visible birds in the flock. foreach item in flock: { count += 1. velocity += (item get-velocity). }
if count == 0: return (0, 0, 0).
velocity /= count.
return velocity - (self get-velocity).
+ to get-center-urge with flock (list): item (object). count (float). center (vector).
# get the average location of all the visible birds in the flock.
foreach item in flock: { count += 1. center += (item get-location). }
if count == 0: return (0, 0, 0).
center /= count.
return center - (self get-location).
+ to check-visibility of item (object): # An item is visible if it is within a certain angle (2.0 radians) # of the direction we're facing (assumed to be a vector in the # direction we're moving).
if (item == self): return 0. if !(item is a "Birds"): return 0. if (self get-angle to item) > 2.0: return 0. if (item check-landed): return 0.
return 1.
+ to get-angle to otherMobile (object): tempVector (vector).
tempVector = (otherMobile get-location) - (self get-location). return angle((self get-velocity), tempVector). }