Evolutionary algorithms optimize civil engineering problems

Evolutionary Algorithms and
Civil Engineering

BY
VAMSIDHAR TANKALA
&
SHREY MODI

DEPARTMENT OF CIVIL ENGINEERING

Consider the following problem

Minimize the function :

f ( x1 , x2 , x3 , x4 )  x  x2  x3  x4
1
2 2 2 2

Various methods

 Mathematical differentiation , and other plotting
techniques.
 A computer program based on these techniques can
be easily formulated.
 What are the issues to be considered ?
- Computational time
- Complexity of the problem – Increase the
parameters and observe the computational time
- Smoothness of the function , if not rugged
techniques work efficiently

Need for Evolutionary Techniques

Vagaries faced by the traditional techniques
 Rugged landscape of the problem
 Presence of many discontinuities
 Simulation of the real world applications where
mathematical formulations are not available :
“BLACK-BOX APPROACHES”
One example : Dynamic traffic simulation.

Need for evolutionary procedures

 “Genetic Algorithms are good at taking
large, potentially huge search spaces and
navigating them, looking for optimal
combinations of things, solutions you
might not otherwise find in a lifetime.”

 - Salvatore Mangano
 Computer Design, May 1995

Brief introduction to GA’s

 Directed search algorithms based on the mechanics of
biological evolution
 Developed by John Holland, University of Michigan
(1970’s)
 To understand the adaptive processes of natural systems
 To design artificial systems software that retains the robustness of
natural systems

 Provide efficient, effective techniques for optimization
and machine learning applications
 Widely-used today in business, scientific and engineering
circles

Genetic Algorithm

 Outline of the steps involved in GA
 Encoding
 Initialization
 Reproduction
 Selection
 Termination Criteria

Deb’s example

Consider a simple can design problem

A cylindrical can considered to have only two
parameters – the diameter d and height h.
Considering that the can needs to have a volume of
atleast 300 ml and the objective of the design is to
minimize the cost of the can material

Objective function

d2 
Minimize f ( d , h)  c (   dh ) 
2 
 d 2h 
Subject to g1 (d , h)   300, 
4 
Variable bounds d min  d  d max 

hmin  h  hmax 

Representing a solution

h (d,h)=(8,10) cm
(chromosome) = 01000 01010

d

Fitness Calculation

F ( s)  0.065[ (8)   (8)(10)]
2

 23.

Fitness – assigning a “goodness” measure

A Sample random generation

Cost -
30

23 30 11

Cost-
40

24 9
37

Selection Operator

 Identify good(usually above-average) solutions in a
population.
 Make multiple copies of good solutions.
 Eliminate bad solutions from the population so that
multiple copies of good solutions can be placed in
the population.

Common Selection methods

 Tournament Selection
 Proportionate selection
 Ranking selection

Tournament selection

Mating Pool 24
23
23 24
37
30

11+30
11+30

24 23 23
24

37 9+40
30
37
9+40 30

Other Selection Operators

 Ranking Selection
 Stochastic Remainder roulette wheel selection
 Proportionate selection

What happens in mating pool??

 Crossover Operation
 Mutation Operation

Crossover operator

23 22
(8,10) 01000 01010 01010 00110 (10,6)
37 39

(14,6) 01110 00110 01100 01010 (12,10)

Mutation Operator

22 (10,6) 01010 00110 01000 00110 (8,6) 16

Overall understanding of GA‟s

Step 1: Encoding Problem
21

 How to encode a solution of the problem
into chromosome ?
 Types of Encoding
 Binary coding 1 0 0 1 1 1 0 1
Difficult to apply directly
Not a natural coding

 Real number coding 2.3352 5.3252 6.2895 4.1525
Mainly for constrained optimization problems

 Integer coding 3 5 1 2 4 8 7 6
For combinatorial optimization problems
Ex. Quadratic Assignment Problems

Step 1: Encoding Problem (Cont.)
22

 Coding Space and Solution Space

Decoding

Coding Space
Genetic Operations Solution Space
Evaluation and Selection

Encoding

23

• Critical issues with encoding
 Feasibility of a chromosome
 solution decoded from a chromosome lies in a feasible region of the
problem
 Legality of a chromosome
 chromosomes represents a solution to a problem
 Uniqueness of mapping (Between Chromosomes and solution to the
problem)
 1 - n mapping (Undesired mapping)
 n – 1 mapping (Undesired mapping) One chromosome
represents only one
 1 – 1 mapping (Desired mapping)
solution to the
problem

24

Coding Space
Solution Space
infeasible one
Feasible space

Coding Space
Solution Space

Step 2: Initialization
25

 Create initial population of solutions
 Randomly

 Local search

 Feasible Solutions

 For optimization problem
 Minimize: F (x1, x2, x3)

 Binary encoding

1 0 1 1 0 0 1 1 1 0 0 1

x1 x2 x3

Step 2: Initialization (Cont.)
26

 Population of solutions
 Fitness of solutions are evaluated (= objective function)

Solution No. Fitness values
1 0 1 0 0 1 0 1 1 0 1 0 1 13.2783
0 1 1 0 1 0 1 0 0 1 0 1 2 20.3749
0 0 1 0 1 0 1 1 1 1 0 0 3 19.8302
Chromosomes

0 1 0 1 0 0 1 0 0 0 1 1 4 52.9405
1 0 0 0 1 0 1 0 1 0 0 1 5 25.8202
1 0 1 1 1 1 0 0 0 0 1 1 6 36.0282
0 0 1 0 1 0 1 1 0 1 1 0 7 70.9202
0 1 1 1 1 0 0 1 1 1 0 1 8 38.9022
0 1 0 1 0 1 0 1 1 0 0 1 9 29.0292
1 0 0 0 1 1 1 1 1 1 0 0 10 21.9292

Step 3: Reproduction
27
 Crossover operation (Based
on crossover probability) Crossover Points

Parent 1
 Select parents from population
based on crossover probability Parent 2
 Randomly select two points
between strings to perform
crossover operation
Offspring 1
 Perform crossover operations
on selected strings Offspring 2
 Known for Local search
operation

Step 3: Reproduction (Cont.)
28

 For the example of optimization problem
 Let the crossover probability be 0.8 Solution Selected
Solution For crossover
No. operation
1 1 0 1 0 0 1 0 1 1 0 1 0 0.9502 > 0.8 NO

2 0 1 1 0 1 0 1 0 0 1 0 1 0.2191 < 0.8 YES

3 0 0 1 0 1 0 1 1 1 1 0 0 0.4607 < 0.8 YES

4 0 1 0 1 0 0 1 0 0 0 1 1 0.6081 < 0.8 YES

5 1 0 0 0 1 0 1 0 1 0 0 1 0.8128 > 0.8 NO

6 1 0 1 1 1 1 0 0 0 0 1 1 0.9256 > 0.8 NO

7 0 0 1 0 1 0 1 1 0 1 1 0 0.7779 < 0.8 YES

0 1 1 1 1 0 0 1 1 1 0 1 0.4596 < 0.8 YES
8
0 1 0 1 0 1 0 1 1 0 0 1 0.9817 > 0.8 NO
9
1 0 0 0 1 1 1 1 1 1 0 0 0.7784 < 0.8 YES
10
Chromosomes Random values [0,1]

29

Solution
Parents Selected Offspring
Selected
2 0 1 1 0 1 0 1 0 0 1 0 1 0 1 1 0 1 0 1 1 0 1 0 1
3 0 0 1 0 1 0 1 1 1 1 0 0 0 0 1 0 1 0 1 0 1 1 0 0

4 0 1 0 1 0 0 1 0 0 0 1 1 0 1 0 1 0 0 1 1 0 1 1 1

7 0 0 1 0 1 0 1 1 0 1 1 0 0 0 1 0 1 0 1 0 0 0 1 0

Crossover Points

8 0 1 1 1 1 0 0 1 1 1 0 1 0 1 0 0 1 0 0 1 1 1 0 1
10 1 0 0 0 1 1 1 1 1 1 0 0 1 0 1 1 1 1 1 1 1 1 0 0

30

 Mutation operation (based on mutation
probability pm)

 each bit of every individual is modified with probability
pm

 main operator for global search (looking at new areas of
the search space)

 pm usually small {0.001,…,0.01}
 rule of thumb pm = 1/no. of bits in chromosome

31

ith solution string from the population
 For optimization
problem 0 0 1 0 1 0 1 1 0 1 0 0

Minimize: F (x1, x2, x3)
 Let pm = 1/12 = 0.083 0.12 0.57 0.62 0.31 0. 01 0.73 0.83 0.63 0.02 0.26 0.94 0.63

 Generate Random
number [0,1] for Mutation
each bit
 Select bits having
probability less than 0 0 1 0 1 0 1 1 0 1 0 0
pm
 Interchange the bits
with each other

Step 4: Selection (“Survival of the fittest”)
32

 Directs the search Population
towards promising (pop_size)
regions in the search Crossover operation
space
Mutation operation
 Basic issues
involved in selection
phase:
 Sampling space:
Parents and Offspring
 Regular sampling space:
all offspring + few parent =
pop_size
Offspring produced

(Cont.)
33

 Basic issues involved in selection phase:
 Sampling space:
 Enlarged sampling space: All offspring + All parent

Crossover operation

Population
(pop_size) Mutation operation

Offspring produced

(Cont.)
34
Selection probability for kth individual
 Sampling
fk
Mechanism: How to pk  pop _ size
select chromosomes  fj
j 1
from sampling space
Based on pk, cumulative probability is
 Basic approaches calculated, and roulette wheel is
constructed
Stochastic Samplings
Roulette Wheel selection: Zone of kth
 To determine survival individual
probability proportional to
the fitness value
 randomly generate a
number between [0,1] and
select the individual

fk is the fitness value of kth individual

(Cont.)
35

 Deterministic Samplings:
select best pop_size individuals from the parents and offspring
No duplication of the individuals

 Mixed Samplings:
both random and deterministic samplings are done

Step 5: Termination Criteria
 Repeating the above steps until the termination criteria is
not satisfied
 Termination criteria
maximum number of generations
no improvement in fitness values for fixed generation

Summary of Genetic Algorithms
36

Begin
{
initialize population;
evaluate population;
while (TerminationCriteriaNotSatisfied)
{
select parents for reproduction;
perform Crossover and mutation;
evaluate population;
}
}

Issues for GA Practitioners
37

 Choosing basic implementation issues:
 Encoding

 Population size, Mutation rate, Crossover rate …..

 Selection, Deletion policies

 Types of Crossover, Mutation operators

 Termination Criteria
 Performance, scalability
 Solution is only as good as the evaluation function
(often hardest part)

Benefits of Genetic Algorithms
38

 Concept is easy to understand
 Modular, separate from application
 Supports multi-objective optimization
 Good for “noisy” environments
 Always an answer; answer gets better with time
 Inherently parallel; easily distributed
 Many ways to speed up and improve a GA-based
application as knowledge about problem domain is
gained
 Easy to exploit previous or alternate solutions
 Flexible building blocks for hybrid applications
 Substantial history and range of use

When to Use a GA
39

 Alternate solutions are too slow or overly
complicated
 Need an exploratory tool to examine new
approaches
 Problem is similar to one that has already been
successfully solved by using a GA
 Want to hybridize with an existing solution
 Benefits of the GA technology meet key problem
requirements

Some GA Application Types
40
Domain Application Types

Control gas pipeline, pole balancing, missile evasion, pursuit

Design semiconductor layout, aircraft design, keyboard configuration,
communication networks

Scheduling manufacturing, facility scheduling, resource allocation

Robotics trajectory planning

Machine Learning designing neural networks, improving classification algorithms, classifier
systems

Signal Processing filter design

Game Playing poker, checkers, prisoner’s dilemma

Combinatorial set covering, travelling salesman, routing, bin packing, graph colouring
and partitioning
Optimization

Sample Applications in Civil Engineering

 Transportation Engineering
 Brief discussion of following areas:
- Dynamic traffic simulation.
- Aggregate blending .
- Back calculation of Pavement Layer Modulii.
 Numerous applications in Structural engineering ,
environmental, geotechnical and water resources
engineering.
Research articles are available in superfluity
concerning applications of GA in civil engineering

Inspiration

 Ants are practically blind but they still manage to find their way
to the food. How do they do it?

 These observations inspired a new type of algorithm called ant
algorithms (or ant systems).

 Result of research on computational intelligence approaches to
combinatorial optimization.

 The algorithm is modeled after the natural behavior of ants.

Natural behavior of ant

Nest Food

Ant search for their food


Nest Food

Obstacle

An obstacle has blocked the path of ants


Nest Food

Obstacle

What to do? Every ant flips a coin and choose a path


Nest Food

Obstacle

Finally, after some time shorter path reinforced

Natural Ants

 Almost Blind.

 Incapable of achieving complex task alone.

 Rely on the phenomena of swarm intelligence for survival.

 Capable of establishing shortest-route paths from their colony to feeding
sources and back.

 Use stigmergic communication via pheromone trails.

Natural Ants

 Follow existing pheromone trails with high probability.

 What emerges is a form of autocatalytic behavior: the more ants follow a
trail, the more attractive that trail becomes for being followed.

 The probability of a path choice increases with the number of times the
same path was chosen before.

Stigmergic

 A term coined by French biologist Pierre-Paul Grasse, is interaction
through the environment.

 Two individuals interact indirectly when one of them modifies the
environment and the other responds to the new environment at a later
time. This is stigmergy.

Stigmergy

Ants uses
stigmergy. PHEROMONES
But how?

Pheromones

These are chemical
What substances dropped by
is us in our path.
Pheromone?

Basic Requirements

 Since the ant algorithms are based on shortest path
finding methodology utilized by the ants in search
for their food, thus their implementation requires:

 The problem to be solved must either be in graphical
format or could be expressed in graphical form.
 Must be finite (i.e. must have a start and end).

Ant Algorithms

 Ant systems are a population based approach. In this
respect it is similar to genetic algorithms.

 Each ant is a simple agent with the following
characteristics:
 It probabilistically chooses the node to visit with certain probability.
 Uses a tabu list to avoid revisit to the node.
 After the completion of tour it lays pheromone trail on each visited
edge.

Flowchart of Ant algorithms

Initialize Evaluate Update
Find Solutions Solutions Pheromone
Ants

Yes Is No Probabilistically find
STOP Termination New solutions based
Criteria met? On pheromone values

Update Evaluate
Pheromone Solutions

Initialization

Find Solutions Solutions Pheromone
Ants


Update Evaluate
Pheromone Solutions

Initialization Initialize
Ants

 Initially ants are randomly placed on the nodes.

 Each edge is initialized with small amount of
pheromones.

 Each edge‟s Visibility, a heuristic value equal to
the inverse of distance between the edge, is
initialized.

Find Solutions

Ants Find Solutions Solutions Pheromone


Update Evaluate
Pheromone Solutions

Find Solutions Find Solutions

 Each ant probabilistically select the next node to visit
with certain probability given by:

Cycle Quantity of pheromone
Number on edge i-j.

1
ij (t )   Distance between
Pij (t )   dij  edge i-j

 1 
 nodesij(t )  dij 
jallowed   α,β constants

Identified
Probability of transition
Using Tabu List
from node i to j

Tabu List

 It is used by the ant to avoid revisit to any node.

 It stores the node to be visited by the ant.

Pheromone Update Update
Pheromone

 After each ant complete their tour, pheromone count
on each edge is updated using:

Pheromone laid by
Evaporation rate each ant that uses
edge (i,j)

Q
 ij (t  1)  (1   ) ij (t )  
kColonythat Lk
used edge ( i , j )

Quantity of
pheromone Total distance traveled
on edge i-j during by ant k during its tour
cycle t+1.

Termination

 The termination criteria commonly used are:

 Designated Maximum number of cycles.
 Specified CPU time limit.

 Maximum number of cycles between two
improvements of the global best solution.

Control Parameters

 Number of ants
 Pheromone Weight ()
 Visibility Weight (β)
 Pheromone persistence (  )
 Number of cycles

Ant Algorithms - Applications

 Travelling Salesman Problem (TSP)
 Facility Layout Problem - which can be shown to
be a Quadratic Assignment Problem (QAP)
 Vehicle Routing

 Stock Cutting (at Nottingham)

ANT COLONY
APPLICATION TO
TRAVELING SALESMAN
PROBLEM – AN EXAMPLE
ILLUSTRATION

Ant Colony Algorithms and TSP

 Ant Colony Optimization was initially designed
for Traveling Salesman Problem.

 At the start of the algorithm one ant is placed in
each city.

 Assuming that the TSP is being represented as a
fully connected graph, each edge has an
intensity of trail on it. This represents the
pheromone trail laid by the ants.


 The distance to the next town, is known as the
visibility, nij, and is defined as 1/dij, where, dij, is
the distance between cities i and j.

 When an ant decides which town to move to
next, it does so with a probability that is based
on the visibility for that city and the amount of
trail intensity on the connecting edge.


 At each cycle pheromone evaporation takes
place.

 The evaporation rate,1- p, is a value between 0
and 1.

 In order to stop ants visiting the same city in the
same tour a data structure, Tabu, is maintained.

Results on TSP with 10 cities

Optimal
Solution

Variants

 Best and Worst Ant System
 The best ant receives reward while the worst ant is punished.
 If the search stucks at a local optimum, restart is employed.
 Maximum and Minimum Ant System
 An upper and lower bound are exposed on the pheromone
levels.
 Search starts using the max.
 Rank Based Ant System
 The ants are sorted wrt. the fitnesses of each tour they find.
 Their pheromone levels are adjusted accordingly
 Elitist Ant System
 The best tour found at each step receives an extra pheromone.

Concluding remarks on Ant algorithms

 Ant algorithms are inspired by real ant colony.

 Probability of ant following certain route is a
function
 Pheromone intensity
 Visibility
 Evaporation

 Ant algorithms are very suitable for problems
having graphical structures.

Inspiration

 It was inspired from the swarms in nature such as
birds, fish, etc.

 PSO algorithm has been originally developed to
imitate the motion of flock of birds.

 Particle Swarm Optimization (PSO) applies concept
of social interaction for problem solving

Particle Swarm Algorithms

 It was developed in 1995 by James Kennedy and Russ Eberhart.

 PSO is a robust stochastic optimization technique based on the movement
and intelligence of swarms.

 In PSO, a swarm of n individuals communicate either directly or indirectly
with one another search directions (gradients).

 It has been applied successfully to a wide variety of search and optimization
problems

PSO Formulation

 The algorithm uses a set of particles flying over a
search space to locate a global optimum.

 A particle encodes a candidate solution to a problem
at hand.

 During an iteration of PSO, each particle updates its
position according to its previous experience and the
experience of its neighbors.

Fundamentals of PSO

 A particle (individual) is composed of:
 Three vectors:
 The x-vector records the current position (location) of the particle in the
search space,
 The p-vector (pbest) records the location of the best solution found so far
by the particle, and
 The v-vector contains a gradient (direction) for which particle will travel in
if undisturbed

PSO: Generic Algorithm Schema

Start

Initialize swarm with random position (x0)
and velocity vectors (v0)

For Each Particle

Evaluate Fitness Next Particle

If
Update Position
fitness(xt)> fitness (gbest)
xt+1= xt+1 + vt+1
gbest=xt

If Update velocity
vt 1  W  vt  c2  rand( 0,1 )  ( pbest  xt )
fitness(xt)> fitness (pbest)
 c3  rand( 0,1 )  ( gbest  xt )]
pbest=xt
gbest= Global Best Position
pbest= Self Best Position If
c1 and c2= Acceleration Coefficients false
Terminate
W = Inertial Weight
true
gbest = output

End

Algorithm Implementation

 The basic concept of PSO lies in accelerating each particle
toward the best position found by it so far (pbest) and the
global best position (gbest) obtained so far by any particle,
with a random weighted acceleration at each time step.

 This is done by simply adding the v-vector to the x-vector to
get another x-vector (Xi = Xi + Vi).

 Once the particle computes the new Xi it then evaluates its
new location. If x-fitness is better than p-fitness, then
pbest = Xi and p-fitness = x-fitness.

Psychosocial compromise
Particle’s best
position so
Particle’ far
s
Current
position pbest Global
x best
gbest position
attained

v
vt 1  W  vt  c1  rand(0, 1)  (pbest  x t )  c2  rand(0, 1)  (gbest  x t )]
gbest = Global Best Position c1 and c2 = Acceleration Coefficients
pbest= Self Best Position W = Inertial Weight

Initial parameters

 Swarm size

 Position of particles.

 Velocity of particles.

 Maximum number of iterations.

Control Parameters

 Swarm size
 Inertial Weight W
 Acceleration Coefficients c1 and c2
 Number of iterations

Inertia Weight W

 A large inertia weight (w) facilitates a global search
while a small inertia weight facilitates a local search.

Larger W Greater Global Search Ability

Smaller W Greater Local Search Ability

Acceleration Coefficients

 Determines the inclination of search.

C1 larger Greater Local Search Ability
than C2

C2 larger Greater Global Search Ability
than C1

Comparison with Evolutionary Algorithms (EAs)

 Unlike EAs, in PSO there is no selection operator.

 PSO does not implement survival of the fittest
strategy and all individuals are kept as members
of the population throughout the course.

Encoding Schema

 Generally PSO is applied over problems involving real
variables.

 However, through the use of proper encoding schema it can
be applied to solve hard combinatorial optimization
problems like Traveling Salesman Problem, Knapsack
Problem, Node Coloring, Sequencing and Scheduling.

Encoding Schema

 For TSP, each particle’s position is coded in the form
of a one dimensional string whose dimensions equals
the number of cities that are to be visited.
 The particles are randomly initialized with rank
vectors or priority numbers.
5 9 2 4 6 3

String
Priority
representation
Numbers
for TSP with 6
cities

Decoding Smallest
Priority
Number
Encoded
String 5 9 2 4 6 3

Decoded
String 1

Least priority is
assigned the first
city

Decoding
Smallest
Priority
Number
Encoded 5 9 N 4 6 3
String

Representing that it
has been decoded

Decoded
String 1 2

Least priority is Repeated till all
assigned the next cities are
city assigned

Decoding

Encoded 5 9 2 4 6 3
String

After Decoding

Finally
Decoded 4 6 1 3 5 2
String

Solution Strategy by Particle Swarm Algorithm

 Randomly initialize the particle‟s position (ranks) and velocity.
 Decode the particles and evaluate objective.
 Store the initial position in particle‟s memory.
 Modify velocity using cognitive and social components and update
position.
 Decode the particles „position and evaluate objective.
 If the position of particle is better than the position stored in
memory, update memory.
 Update the global best if a better particle is obtained.
 Repeat the process till required no. of iterations are complete.
 The particle with best position is the output.

Results of PSO on TSP with 10 Nodes

Results of PSO on TSP with 10 Nodes

PSO took
relatively large
time to evolve
the optimal
solution

Concluding remarks on “Particle Swarm”

 Fast convergence thus time requirement is less.
 Global as well as Local search component.
 Dependence on parameter tuning is less.
 More effective on problems involving real values.
 Chances of early convergence due to high
convergence speed.

Artificial Immune Systems

 AIS are adaptive systems inspired by theoretical immunology
and observed immune functions, principles and models, which
are applied to complex problem domains (de Castro and Timmis)
 A recently developed evolutionary technique inspired by theory
of Immunology

A way to study the response of immune system,
when a non-self Antigen pattern is recognized
by Antibody

Biological Immune System
Efficiency of the acquired response depends upon the ability
of antibodies to recognize the antigens, depends upon

1016 Antigens for
less than 100
antibody genes

o Generalization
o Screening
Self/Non-self
o Memory
Discriminatio
n

Ability to
remember
previous
infections

Artificial Immune System
 History of Artificial Immune System
Initially developed from the theory of “immunology”
in mid 1980’s
In 1990, first use of immune algorithm to solve
optimization problem
In mid 1990: Application to Computer Security
In mid 1990: Machine Learning

 Artificial Immune System: An Optimization View

Objective
Entire Solution Functions
Building Blocks Constraints
Feasible Solutions

Artificial Immune Systems

 Basic Elements

 Immune Systems: To protect the body from the
foreign matters
 Antigen: Any foreign disease causing elements
 Antibody: Utilized to identify, bind and eliminate
antigens

General Framework for AIS- The AIS
Cycle

Population
Selection
Initialization

Cloning &
Evaluation
Hypermutation

 AIS: A Generic Framework
Immune
Algorithm

Affinity
Measures

Representation

Application
Domain

Flow of the Algorithm

Hypermutation
of each clone

Population P
of individuals
Clone Pool of
the population

Repository of
Probabilistically select P
best individual good solution

When search gets stagnated good solutions are sent to
the current population

 Artificial Immune System: An Assessment
 Advantages
General Purpose AIS tools
Easily Extensible
Potential for distribution
 Disadvantages
Parameter Sensitive
Computationally Expensive

 Distinctive Features & Their Applications:

Features Applications

Learning & Adaptation Security

Immunological Memory Pattern Recognition

Self/Non-self Classification Heuristic Optimization

Self Organizing Modeling & Agents Application

Localization & Circulation Clustering

Autonomous/Decentralized Concept Learning & Recommender System


 AIS: Potential Area of optimization
 Fault & Anomaly Detection
 Data Mining (Machine Learning, patter recognition)
 Agent Based systems
 Autonomous Control
 Information Security System
 Scheduling

Dynamic traffic simulation

 CALIBARTION OF MESOSCOPIC TRAFFIC
SIMULATION USING POPULATION BASED
EVOLUTIONARY ALGORITHMS
- methodology to calibrate dynamic traffic
simulation models with real data acquired from
traffic counts and travel time measurements
acquired from GPS devices

Brief outlook

 To use a tool called METROPOLIS
 No Mathematical function involved, hence a need for
simulation arises – simulation of the real world
conditions.
 A simulation of real time traffic will be processed in
the model and it gives different indicators as output.
One of the indicator is the travel time along the
defined paths in a network
 Initially ,tested on toy networks (network containing
small networks)

Computational Details

 Programmed in the following way :
- A GUI platform developed in java which works like
a compiler for optimization
- The main features of the compiler are :
* Any EA can be embedded
* Any problem can be optimized

ALGORITHM OPTIMISATION PROBLEM

PLATFORM

Overall framework

NODE-1

RANDOMLY CALCULATE
GENERATE FITNESS
PLATFORM NODE-2
TRAFFIC
VARIABLES FITNESS

NODE-3
R
E
S NODE-4
U
L
T
S

Evolutionary algorithms optimize civil engineering problems

Empfohlen

Empfohlen

Weitere ähnliche Inhalte

Was ist angesagt?

Was ist angesagt? (20)

Andere mochten auch

Andere mochten auch (6)

Ähnlich wie Evolutionary algorithms optimize civil engineering problems

Ähnlich wie Evolutionary algorithms optimize civil engineering problems (20)

Kürzlich hochgeladen

Kürzlich hochgeladen (20)

Evolutionary algorithms optimize civil engineering problems