Transcript Biology: flocking, herding & schooling Day 5 COLQ 201 Multiagent modeling Harry Howard Tulane University Course organization  http://www.tulane.edu/~howard/Multiagent/  Photos? 22-Jan-2010 COLQ 201, Prof.

Biology: flocking,
herding & schooling
Day 5
COLQ 201
Multiagent modeling
Harry Howard
Tulane University
Course organization
 http://www.tulane.edu/~howard/Multiagent/
 Photos?
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
2
Photos
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
3
Boids
http://www.red3d.com/cwr/boids/
What did you learn
about Boids?
 Date?
 First appearance?
 Movies?
 A-life?
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
5
Steering behaviors
 They describe how an individual boid
maneuvers based on the positions and
velocities its nearby flockmates:
separation
alignment
cohesion
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
6
Separation
 Steer to avoid
crowding local
flockmates.
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
7
Alignment
 Steer towards the
average heading of
local flockmates.
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
8
Cohesion
 Steer to move toward
the average position of
local flockmates.
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
9
Neighborhood
 Distance (measured from
the center of the boid) and
 Angle, measured from the
boid's direction of flight.
 It could be considered a
model of
 limited perception (as by
fish in murky water)
 the region in which
flockmates influence a
boid's steering.
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
10
MyFlocking
Community model
Overview
 What do you see in the interface?
 How does it compare to Boids?
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
12
Questions
 Keeping the other parameters at their default
values (vision = 3, min-separation = 1, max-alignturn = 5, max-cohere-turn = 3, max-separate-turn
= 1.5), …
 what does vision do?
 what does minimum-separation do?
 what does max-align-turn do?
 what does max-cohere-turn do?
 what does max-separate-turn do?
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
13
Conclusions
Chaos and emergence
 In Boids (and related systems) interaction between
simple behaviors of individuals produce complex
yet organized group behavior.
 The component behaviors are inherently nonlinear, so
mixing them gives the emergent group dynamics a
chaotic aspect.
 At the same time, the negative feedback provided by
the behavioral controllers tends to keep the group
dynamics ordered.
 The result is life-like group behavior.
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
15
Time scales
 A significant property of life-like behavior is
unpredictability over moderate time scales.
 At very short time scales, the motion is quite
predictable: one second from now a boid will be
traveling in approximately the same direction.
 Yet if the boids are flying primarily from left to right, it
would be all but impossible to predict which direction
they will be moving (say) five minutes later.
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
16
At the edge of chaos
 This property is unique to complex systems and
contrasts with both random behavior (which has
neither short nor long term predictability) and
ordered behavior (which is predictable in both the
short and long term).
 This fits with Langton's 1990 observation that lifelike phenomena exist at the edge of chaos.
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
17
Chaos
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
18
Agents
 Boids is an example of an individual-based
model, a class of simulation used to capture
the global behavior of a large number of
interacting autonomous agents.
 Individual-based models are being used in
biology, ecology, economics and other
fields of study (and in this course).
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
19
Complexity
 A straightforward implementation of the boids
algorithm has an asymptotic complexity of O(n2).
 Each boid needs to consider every other boid, if only to
determine whether it is a nearby flockmate.
 However it is possible to pare this cost down to nearly
O(n) by the use of a suitable spatial data structure
which allows the boids to be kept sorted by their
location.
 Finding the nearby flockmates of a given boid then
requires examining only the portion of the flock which
is within the general vicinity.
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
20
Programming NetLogo
The NetLogo world
 … is a two dimensional world that is made
up of turtles, patches and an observer.
 The patches create the ground in which the
turtles can move around on and
 the observer is a being that oversees
everything that is going on in the world.
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
22
P1
 I will ask you to open a model that you have
not seen, and I will ask you to answer some
questions about it and how it works.
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
23
Next time
 P1
 Biology: from foraging to graph theory
Ants2, AntSystem
22-Jan-2010
COLQ 201, Prof. Howard, Tulane University
24