lecture 9 aco

37
Department of Biomedical, Industrial and Human Factors Engineering 1 Ant Colony Systems and the Ant Algorithm

Upload: mcradc

Post on 18-May-2015

1.320 views

Category:

Technology


0 download

TRANSCRIPT

Page 1: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering1

Ant Colony Systems

and the

Ant Algorithm

Page 2: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering2

REAL ANT BEHAVIOR

Page 3: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering3

Natural behavior of ants have inspired scientists to mimic insect

operational methods to solve real-life complex problems

By observing ant behavior, scientists have begun to understand

their means of communication

Ant-based behavioral patterns to address combinatorial

problems - first proposed by Marco Dorigo

REAL ANT BEHAVIOR

Ants secrete pheromone while

traveling from the nest to food,

and vice versa in order to

communicate with one another to

find the shortest path

Page 4: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering4

EXPERIMENTAL STUDY OF ANTS

The more ants follow a trail, the more attractive that trail

becomes for being followed

NEST FOODNEST FOODNEST FOOD

Page 5: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering5

ANT Behavior

The more ants follow a trail, the more attractive that trail

becomes for being followed

Page 6: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering6

ANT Behavior

Even when the tracks are equal the behavior will encourage one

over the other--convergence (Deneubourg et al)

Page 7: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering7

ROUTE SELECTION

Ants are forced to decide whether they should go left or right, and

the choice that is made is a random decision

Pheromone accumulation is faster on the shorter path

The difference in pheromone content between the two paths over

time makes the ants choose the shorter path

Positive feedback mechanism to arrive at the shortest route while

foraging

Stygmergy or stigmergetic model of communication

Different optimization problems have been explored using a

simulation of this real ant behavior

Page 8: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering8

TRAVELING SALESMAN PROBLEM

Page 9: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering9

PROBLEM DEFINITION

OBJECTIVE

Given a set of n cities, the Traveling

Salesman Problem requires a salesman

to find the shortest route between the

given cities and return to the starting

city, while keeping in mind that each city

can be visited only once

Page 10: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering10

WHY IS TSP DIFFICULT TO SOLVE?

Finding the best solution may entail an exhaustive search for all

combinations of cities. This can be prohibitive as “n” gets very

large

Heuristics like a “greedy” route doesn’t guarantee optimal

solutions

d

e

gfb

c

h

a d

e

gfb

c

h

a

Page 11: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering11

TSP Applications

• Lots of practical applications

• Routing such as in trucking, delivery, UAVs

• Manufacturing routing such as movement of parts along manufacturing floor or the amount of solder on circuit board

• Network design such as determining the amount of cabling required

• Two main types

– Symmetric

– Asymmetric

Page 12: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering12

General Formulation - Symmetric

iji ij ijxd∑∑>min

ixxij ij jiij ∀≥+∑∑< > 2,,

32,, ≥∋∀≥+∑∑∑∑∈>∉ ∉>∈SSxxSi ijSj Si ijSj

jiji

+∈Bxji,

Page 13: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering13

General Formulation - Asymmetric

iji ij ijxd∑∑≠min

ixjij ∀=∑1,

ixiji ∀=∑1,

21, ≥∋∀≥∑∑∈∉ SSxSi Sj ji

+∈Bxji,

Page 14: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering14

TSP Heuristics

• Variety of heuristics used to solve the TSP

• The TSP is not only theoretically difficult it is also difficult in practical application since the tour breaking contraints get quite numerous

• As a result there have been a variety of methods proposed for the TSP

• Nearest Neighbor is a typical greedy approach

Page 15: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering15

Simple Examples

2 3 4 5 6 7 8 9 10

1 Chicago 96 105 50 41 86 46 29 56 70

2 Dallas 78 49 94 21 64 63 41 37

3 Denver 60 84 61 54 86 76 51

4 KC 45 35 20 26 17 18

5 Minn 80 36 55 59 64

6 OK City 46 50 28 8

7 Omaha 45 37 30

8 St Louis 21 45

9 Sprngfld 25

10 Wichita

1

28

3

9

7

6

5

4

6

21 2

6

22 2

2 1

621

2

2

Page 16: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering16

Nearest Neighbor Solution

Iteration Node Arc Cost Total

1 1

2 8 1,8 29

3 9 8,9 21 50

4 4 9,4 17 67

5 10 4,10 18 85

6 6 10,6 8 93

7 2 6,2 21 114

8 7 2,7 64 178

9 5 7,5 36 214

10 3 5,3 84 298

11 3,1 105 403

Page 17: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering17

0

50

100

150

200

250

300

350

400

450

500

0 50 100 150 200 250 300 350 400 450 500

Way Point

Depot

Obj Fx: min d

Larger TSP Example

Page 18: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering18

0

50

100

150

200

250

300

350

400

450

500

0 50 100 150 200 250 300 350 400 450 500

Way Point

Depot

d=3138

Initial Order Solution

Page 19: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering19

0

50

100

150

200

250

300

350

400

450

500

0 50 100 150 200 250 300 350 400 450 500

Way Point

Depot

d=2108

Nearest Neighbor Solution

Page 20: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering20

0

50

100

150

200

250

300

350

400

450

500

0 50 100 150 200 250 300 350 400 450 500

Way Point

Depot

d=1830

Tabu Search Solution

Page 21: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering21

THE ANT COLONY OPTIMIZATION METAHEURISTIC

Page 22: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering22

GOAL OF ACO HEURISTIC

Artificial ants form a multi-agent system performing the

functions as observed in the real ant system

Exploit stigmergistic communication

The ACO meta-heuristic relies on the co-operation of a

group of artificial ants to obtain a good solution to a

discrete optimization problem such as the TSP

Artificial ants are mutants of a real ant system

The resulting shortest route mapping determined by the agents

can be applied to the optimization problem

Page 23: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering23

ACO CHARACTERISTICS

Exploit a positive feedback mechanism

Demonstrate a distributed computational architecture

Exploit a global data structure that changes dynamically as each ant transverses the route

Has an element of distributed computation to it involving the population of ants

Involves probabilistic transitions among states or rather between nodes

Page 24: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering24

REAL vs. ARTIFICIAL ANTS

• Discrete time steps

• Memory Allocation

• Quality of Solution

• Time of Pheromone deposition

• Distance Estimation

REAL ANT ARTIFICIAL ANT

Page 25: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering25

FLOWCHART OF ACO

Have all cities been

visited

Have the maximum

Iterations been performed

START ACO

Locate ants randomly in cities across the grid and store the

current city in a tabu list

Determine probabilistically as to which city to visit next

Move to next city and place this city in the

tabu list

Record the length of tour and clear tabu list

Determine the shortest tour till now and

update pheromone

NO

YES

STOPACO

YESNO

Page 26: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering26

KEY PARAMETERS

Trail intensity is given by value of ij which indicates the intensity of the pheromone on the trail segment, (ij)

Trail visibility is ij = 1/dij

The importance of the intensity in the probabilistic transition is

The importance of the visibility of the trail segment is

The trail persistence or evaporation rate is given as

Q is a constant and the amount of peromone laid on a trail segment employed by an Ant; this amount may be modified

in various manners

Page 27: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering27

PROBABILISTIC CITY SELECTION

Helps determine the city to visit next while the ant is in a tour

Determined by variables such as the pheromone content in an

edge (i,j) at time instant t, heuristic function of the desirability of

adding edge, and their control parameters

[ ] [ ][ ] [ ]

)(

)(

0

)(

)(

)()(

iJjf

iJjift

t

tp

k

kilil

ijij

kij

ikJl

⎪⎪⎩

⎪⎪⎨

= ∑∈

βα

βα

ητ

ητ

Page 28: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering28

PHEROMONE UPDATING

Using the tour length for the k-th Ant, Lk, the quantity of pheromone

added to each edge belonging to the completed tour is given by

( ) ( )( )tTjiedgeif

tTjiedgewhereL

Qt

k

k

k

kij ∉

⎪⎩

⎪⎨⎧

=Δ),(

),(

)()()1()1( ttt ijijij Δ+−=+

The pheromone decay in each edge of a tour is given by

Page 29: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering29

ACTUALLY 3 ALGORITHMS

The ant-cycle is the approach discussed so far

Information is updated at the end of each tour as such function of tour length

The ant-density is an approach wherein the pheromone quantity Q is deposited once the segment is transversed

Pretty much a greedy approach (local information) and not really providing relative information

The ant-quantity is an approach wherein the pheromone quantity Q/dij is deposited once the segment is transversed

Also a greedy approach but providing some relative information by scaling Q by the length of the segment

Page 30: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering30

Consider the Case Studies in Papers

Page 31: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering31

EXTENSIONS

Found that communication among the ants via the intensity factor is important; makes sense

since it provides some global insight

Found that a good number of ants is about equal to the number of cities

Found that the initial distribution of the ants among the cities does not really matter

Found that an elitist strategy in which the segments on the best solution(s) is(are) continually

reinforced work well so long as there were not too few or too many elitist solutions

Page 32: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering32

APPLICATIONS

Page 33: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering33

APPLICATIONS

Page 34: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering34

Population-Based Incremental Learning

Lots of similarities to the ACO

Actually inspired by genetic algorithms

Generate members of a population randomly based on probability of selection functions that are

increased or decreased based on the quality of past solutions involving the member variables

Once a population is generated, evaluate and then increase or lower the probability used by the

generating vector to encourage better solutions.

Page 35: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering35

Population-Based Incremental Learning

Benefits

Will converge to solutions under correct circumstances

Efficient in terms of storage

Computationally pretty cheap

Involves learning

Disadvantages

Keeps primarily a local focus

Cannot handle interdependence among parameters very well

Will need to involve penalty functions for constraints

Page 36: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering36

Population-Based Incremental Learning

A solution to overcome disadvantages proposed by Miagkiky

and Punch

Combine reinforcement with population generation

Page 37: Lecture 9 aco

Department of Biomedical, Industrial and Human Factors Engineering37

QUESTIONS??