Vous êtes sur la page 1sur 21

Presented By:

Ant Colony Aditya


Group 5
(PGP/21/191)

Optimization
Gireeshkumar (PGP/21/204)
Prabhumathi M (PGP/21/217)
Sai Shyam T M (PGP/21/230)
Vibhore Sharma (PGP/21/243)
 Ants choose the
path which has
Introduction highest amount
of pheromone.

 Originated
from the
study of ant  Pheromones released by ants
colonies. while walking from nest to
destination (food) help other
ants to follow the same path.
 Marco Dorigo (Italy)
was the first one to
publish ACO in 1992.
 Ants produce a chemical
named as Pheromones
 ACO is a meta- which is used to
heuristic and communicate among
probabilistic them.
technique.
Each ant decides on whether to take a
particular path based on probability
Where to go? given by the probability computing
equation.

150

180
??
Natural Behavior of Ants
Choosing shortest path
Probability computation
Formula for ACO:

The probability of ant going from (𝜏𝛼 𝑖 𝑗 )(𝜂𝛽 𝑖 𝑗 )


node i to node j is given by, 𝑝𝑖 , 𝑗 = 𝛼
, ,
∑(𝜏 𝑖 𝑗 )(𝜂𝛽 𝑖 𝑗 )
, ,
Where,
𝜏 𝑖, 𝑗 is the amount of pheromone on edge i, j  A relatively shorter path will comparatively
have more pheromone deposited on the
𝛼 is a parameter to control the influence of 𝜏 𝑖, 𝑗 return trip
1
𝜂 𝑖, 𝑗 is the desirability of edge i, j (typically ൗ𝑑 )
𝑖, 𝑗

𝛽 is a parameter to control the influence of 𝜂 𝑖 , 𝑗


Pheromone update equation,
𝜏 𝑖, 𝑗=(1-𝜌)𝜏 𝑖, 𝑗 + ∆𝜏 𝑖, 𝑗
𝜏 𝑖, 𝑗 is the amount of pheromone on a given edge i, j
𝜌 is the rate of pheromone evaporation
∆𝜏 𝑖, 𝑗 is the amount of pheromone deposited, typically given by

1ൗ 𝑘 If ant k travels on edge i,j


∆𝜏𝑘 𝐿
𝑖, 𝑗=
0 otherwise

where 𝐿𝑘 is the cost of the kth ant’s tour (typically length).


Pheromone values are updated by all the ants that have completed the tour,

𝜏 𝑖 , 𝑗 ← (1−𝜌)𝜏 𝑖, 𝑗+∑𝑚 𝑘
𝑘=1 ∆𝜏 𝑖 , 𝑗

m is the number of ants


∆𝜏𝑘 𝑖, 𝑗 is pheromone quantity laid on edge (i, j) by the kth ant
Cont…

𝜏 𝑖, 𝑗=(1-𝜌)𝜏 𝑖 , 𝑗 + ∆𝜏 𝑖, 𝑗best

Where,

1ൗ 𝑏𝑒𝑠𝑡 If best ant k travels on edge i,j


𝜏 𝑖, 𝑗 best= 𝐿
0 otherwise

𝐿𝑏𝑒𝑠𝑡 can be set to the length of the best tour found in the current iteration or the best solution found
since the start of the algorithm.

 Min and Max values of the pheromone are explicitly limited,


 𝜏 𝑖, 𝑗 is constrained between 𝜏 max and 𝜏 min(explicitly set by algorithm designer).
 After pheromone update, 𝜏 𝑖, 𝑗 is set to max if 𝜏 𝑖 , 𝑗 > 𝜏 max and to min if 𝜏 𝑖, 𝑗 < 𝜏 min
Travelling
Salesman
Problem
Using Ant Colony Optimization
ACO Algorithm

 Initiation of all paths (uniform pheromone


level)
 Random placement of ants
 Start forward movement, probabilistically
select next node by relative pheromone
levels of surrounding nodes
 Ensure loops are not formed
 Backtrack the steps
 According to the evaporation rate ρ, update
the overall trail
 For the retraced ants, update the change in
amount of pheromone Δτij
 Exit
Example
1

D E

B
   
  ij (t )   ij 
if k  allowed k

How does the ant choose k
   
pij t      ij (t )  ij
 

node?  kallowedk
 0 otherwise
ηAj  Initially the node is at node “A” and the
PAj = η
AB +ηAC + ηAD +ηAE+ηAF possible paths to cover are AB, AC, AD,
1
150
AE,AF
PAB = 1 1 1 1 1 = 0.1875
+ + + +
150 100 120 180 200  Apply the probability formula with α=β=1
1
PAC = 100
= 0.2812  ηij is calculated by the formulae ηij = 1/Dij,
1 1 1 1 1
+ + + +
100 150 120 180 200 where Dij is the distance between nodes i
1 and j.
120
PAD = 1 1 1 1 1 = 0.2343
+ + + +
120 100 150 180 200  DAB =150; DAC = 100; DAD=120; DAE =180; DAF
1 =200
180
PAE = 1 1 1 1 1 = 0.1562
+ + + +
180 100 120 150 200
 Therefore,
1
200
ηAB = 1/150; ηAC = 1/100 ; ηAD = 1/120;
PAF = 1 1 1 1 1 = 0.1456
+ + + +
180 100 120 150 200
 ηAE = 1/180; ηAF = 1/200
STEP – 1
3
1

6
C
A

2 F

5
B 4

E
D
STEP – 2
6
2

3
F-C
B-A

1 C-F

4
A-B 5

D-E
E-D
STEP – 3
2
6

B-A-C 4

F-C-A

5
D-E-F
3

E-D-B 1

A-B-D C-F-E
STEP – 4
4
5

2
D-E-F-C
E-D-B-A

6 B-A-C-F
1
F-C-A-B 3

C-F-E-D A-B-D-E
STEP – 5
5
4

E-D-B-A-C 1

D-E-F-C-A

3 A-B-D-E-F

C-F-E-D-B 6

F-C-A-B-D B-A-C-F-E
STEP – 6
1
3

5
A-B-D-E-F-C
C-F-E-D-B-A

4
E-D-B-A-C-F

6
D-E-F-C-A-B 2

F-C-A-B-D-E
B-A-C-F-E-D
Practical applications

 Water distribution system (WDS)


 Network of components that transport water from
a source to the consumers

 Google Maps
 Tourist attractions have suffered heavy traffic problems during vacations or holidays
 ACO route planning model was used in order to establish a dynamic planning system that can
instantaneously provide optimal route information
 Google Maps used ACO to assist in tailoring travel routes for each visitor
 The results show that the mean total desired value of improved ACO algorithm is 25.53% higher;
the average travel time is reduced by 34.32%
Disadvantages

Advantages &  Theoretical analysis is difficult:


Disadvantages  Due to sequences of random decisions
(not independent)
 Probability distribution changes by
iteration
Advantages  Research is experimental rather than
theoretical
 It is versatile – i.e. it can be applied to  Convergence is guaranteed, but time
similar versions of the same problem to convergence uncertain
 It is robust – i.e. it can be applied with  Coding is complicated
minimal changes
 It is a population based model – it
allows exploitation of positive
!!!!!
feedback as a search mechanism
Thank You!!

Vous aimerez peut-être aussi