Download display

Multidisciplinary Senior Design
Project Readiness Package
Project Title:
Noise Cancellation using an Evolutionary Algorithm (EA) Phased
Array Technique
Project Number:
(assigned by MSD)
Primary Customer:
(provide name, phone
number, and email)
Sponsor(s):
(provide name, phone
number, email, and
amount of support)
RIT
Dr. Vincent J Amuso
Tel: 585.506.2405
Email: [email protected]
RIT
Preferred Start Term:
Fall 2016
Faculty Champion:
(provide name and
email)
Dr. Vincent J. Amuso
[email protected]
Other Support:
Project Guide:
(assigned by MSD)
Dr. Vincent J. Amuso
Prepared By
8/1/2016
Date
Received By
Date
Items marked with a * are required, and items marked with a † are preferred if available, but we
can work with the proposer on these.
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
Project Readiness Package
Template Revised Spring 2016
Project Information

Overview:
Noisy office environments often make it difficult for employees to concentrate. The
noise is characterized by a combination of both stationary and non-stationary noise and
interference sources. This makes traditional noise cancelling techniques very difficult to
implement in real or near real-time. In this project we propose using a network of acoustic
receivers (microphones) and transmitters (speakers) that are resident in laptop and
desktop computers located in the aforementioned cubicles of a multi dwelling office
environment to gather acoustic signals and “evolve” a noise cancelling system.
The architecture of the proposed system is based on phased array radar and sonar
technology. This technique was originally developed for use during World War II (see
http://www.radartutorial.eu/06.antennas/Phased%20Array%20Antenna.en.html for a
description). Through the years that followed the technology has been adapted for many
commercial applications. As compared to conventional mechanically-scanned systems
performing the same functions, phased array systems provide superior capabilities in
terms of data rates, scanning ability, and the ability to both receive and transmit
sophisticated array patterns.
* Preliminary Customer
Requirements (CR):
The proposed design of the Noise Cancellation System using an Evolutionary Algorithm
(EA) Phased Array Technique will be based on the following requirements:
1.
2.
3.
4.
5.
The system should use laptops and/or desktops that are resident in the work environment
The system should require minimal adjustment by the user(s)
The system bandwidth should be the audible range of an average adult
The system should respond in near real-time
The system should significantly reduce the noise environment
† Functional
Decomposition:
Due to the anticipated design requirements a traditional closed form design solution
would be extremely difficult if not impossible to achieve. After the requirements are determined
and incorporated into hardware/system specifications, a technique that lends itself to the
optimization of multiple objectives with a multi- dimensional design parameter space should be
employed. Evolutionary algorithms are ideally suited to this type of problem.
Figure 1 below is a flowchart of a similar system of the one being proposed in this
project. The system below would need to be modified but the architecture closely mimics the one
needed to achieve the customer requirements.
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
Project Readiness Package
Template Revised Spring 2016
Trade Space Analyses
Choose
Technology
Options
(i.e.Optical,
MEMS,etc.)
Air Force
Requirements
(see below)
1.
2.
3.
4.
5.
6.
7.
8.
Extract
Technology
Attributes
Generate
Objective
Functions
Evolutionary
Multi-objective
Optimization
(EMO) Algorithm
(see 2.1.4)
Generate
Design
Specs
Active power distribution
Reconfigurable variable
gain power
splitters/combiners
Bandwidth 0.5 – 18 GHz
Scan a minimum of +/60°
Resolution of 7.5° max in
at least one plane
Minimum 16 elements
array size with ability to
scale to a larger
numbers of elements
Thickness no greater
than 1”
Weight no more than
1lbs
Design
Hardware
Build
Prototype
Generate
Test Plan
Test
Prototype
1.
2.
Tested Hardware
Novel EMO based
design technique
Fig. 1. Example of a proposed project flowchart taken from phased array radar system
The proposed approach uses Evolutionary Algorithms that employ the concepts of “the
survival of the fittest”, and mathematically implement this concept into an algorithm in order to
produce a generic stochastic approach to solving single or Multi-objective Optimization
Problems (MOP’s). A simple flowchart that illustrates the basic EA structure is illustrated in
Fig. 2.
Initialize Population
Determine Fitness
Create Offspring
N
Apply Selection
Terminate?
Y
Fig. 2. Flowchart of a typical Evolutionary Algorithm
*
Preliminary Engineering Requirements (ER):
The Engineering Requirements will be determined by the MSD team and Dr. Amuso.
The requirements may well be determined by the expertise and mix of the MSD team members.
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
Project Readiness Package
Template Revised Spring 2016
Metrics:
TBD
Specifications:
TBD
Constraints:
The MSD team will use the SPEA2 algorithm implemented by Dr. Amuso and his graduate
students. The algorithm is coded in MATLAB. Since the students will be constrained to use the
SPEA2 algorithm which is an Evolutionary Multi-objective Optimization algorithm, a brief
synopsis of this type of algorithm is provided.

In order to fully explain an Evolutionary Multi-objective Optimization algorithm, several
concepts will be defined and clarified:
1.
2.
3.
4.
The Multi-objective Optimization Problem (MOP)
Pareto optimality
Mathematical encoding of the population
The Strength Pareto Evolutionary Algorithm (SPEA2)
(1) The Multi-objective Optimization Problem (MOP) can be defined as the problem of
finding a vector of decision variables which satisfies constraints and optimizes a vector function
whose elements represent the objective functions [10]. These functions form a mathematical
description of performance criteria which are usually in conflict with each other. Hence, the term
“optimize” means finding a solution which would give the values of all the objective functions
acceptable to the decision maker.
An illustration of the mapping of solutions, or decision space, to a multi-objective
function space is given in Fig. 3 below.
Objective Function Space
Decision Variable Space

Feasible Region   x
R
n

  y  R
x2
k

F3
x1

x  x , x ,, x
1
2
n
F2
F1

T




f x    f x , f x ,, f x 
T
1
2
k
Fig. 3. A mapping illustration from decision variable space to multi-objective
function space
A solution to an MOP can be described as follows. Find the vector



*
*
*T
x*  x1 , x 2 ,  , x n
which satisfies the m inequality constraints
g i x   0 for i  1,2,...m
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
(1)
(2)
Project Readiness Package
Template Revised Spring 2016
and the p equality constraints
hi x   0 for i  1,2,...p
and optimizes the vector function
(3)


 


 T
f x   f1 x , f 2 x ,  f 2 x 
.
(4)
The objective functions used to represent the particular sensor parameters are represented

by the f x  vector.
Let

O i 
O i 
O i 

xO i   x1 , x2 ,, xn

T
(5)
be a vector that optimizes the ith objective function. The vector

x O i   
(6)
is such that
 
 O i 

fi x
 opt fi x 
xΩ
The vector


f  f1O , f 2O ,  f kO

T
(7)
(8)
is ideal for an MOP and the point in Rn space that determines this vector is called the ideal
vector.
(2) Pareto Optimality The existence of the ideal vector is extremely rare especially as the
dimensionality of the fitness vector increases. Therefore optimality must be defined in such a
way as to satisfy a compromise of the optimality of individual objectives to be optimized. The
idea of a trade off optimality criteria was originally proposed by Francis Ysidro Edgeworth in
1881 and generalized by Vilfredo Pareto in 1896. It is generally referred to as Pareto optimum. A
solution is Pareto Optimal if there exists no feasible vector which would decrease some criterion
without causing a simultaneous increase in at least one other criterion [9]. Mathematically this is
represented as follows:


A point x *   is Pareto Optimal if for every x   and I  1,2,...k  either

 f x   f  x *  
i  I  i
i   
(9)
or there is at least one i  I such that


f x   f  x *  .
i
i 
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
(10)
Project Readiness Package
Template Revised Spring 2016
(3) Mathematical Encoding of the Population In order to implement the algorithm, the
population members, which represent possible solutions to the problem, must be represented in a
tangible way so that they may be operated upon by the mathematical representations of the
evolutionary principles of selection and adaptation. The individuals are represented by a
collection of genes that when grouped together form chromosomes. These chromosomes contain
all the information required to represent and distinguish the difference amongst individual
population members. From an algorithmic point of view the individual is nothing more than an
encoded solution to a mathematical problem. Using a human as an example to illustrate the
aforementioned concepts, the nucleus of most human cells is composed of two sets of
chromosomes, one set given by each parent. Each set has 23 single chromosomes. The
chromosomes contain genes which contain specific information pertaining to an individual.
Using eye color as an example, the gene that determines this trait is found in the same
chromosome and position for all humans. The size of the encoding needed to represent the
information is the same for all humans as well. The actual values that are contained in those
physical locations (a sequence of A T G or C) represent the actual color (blue, brown green etc.)
for that individual. The actual values are called alleles. Returning to the EA representation, the
mathematical encoding used is either binary encoding or real number encoding. Using the human
eye color example in a binary encoding scheme, two positions in a gene could be mapped to the
actual eye color using a two bit binary representation such as 00 –Blue, 01 –Green, 10 –Brown,
11 –Hazel.
(4) The Strength Pareto Evolutionary Algorithm (SPEA2).Zitzler & Thiele have developed
an approach to multi-objective optimization, named the Strength Pareto Evolutionary Algorithm
2 (SPEA2) . The SPEA2 uses a mixture of established and new techniques in order to find
multiple Pareto-optimal solutions in parallel. The SPEA2 algorithm incorporates a fine-grained
fitness assignment strategy that incorporates density information, a density estimation technique
that is an adaptation of the kth nearest neighbor method, and an enhanced archive truncation
method that guarantees the preservation of boundary solutions. This truncation method preserves
boundary points on the Pareto front. The two main improvements of SPEA2 over SPEA are 1)
a more precisely guided search process fostered by a nearest neighbor density technique,
and 2) an archive truncation method that guarantees the preservation of boundary
solutions to ensure that the entire span of the Pareto front remains intact.
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
Project Readiness Package
Template Revised Spring 2016
Generate Initial Population
P0 and an Empty Archive
A0. Set t = 0
Calculate Fitness of
Individuals in Pt and At
Perform Environmental
Selection to Fill At+1
Too
Small
Add Best Dominated
Members of Pt
Too
Large
Truncate Using Nearest
Neighbor Criteria
Check Size of At+1
t = max generation?
Y
Terminate, At+1
Contains Pareto Front
N
Perform Mating Selection,
Crossover & Mutation using only
At+1 to form Pt+1. Set t = t +1
Fig. 4. SPEA2 Flowchart
Figure 4 depicts the flowchart for SPEA2. In SPEA2, each individual’s fitness is based upon the
solutions that dominate it as well as the solutions that it dominates. For each individual i, a
strength value S(i) is determined that represents the number of solutions that it dominates. In
mathematical terms,
S (i )   j | j  Pt  At  i j 
(11)
where | | denotes the cardinality of a set and + represents a multi-set union. These strength
values are then used to assign a raw fitness, R(i), to each individual.
R(i)   S ( j )
jPt  At , j i
(12)
Thus, the raw fitness of an individual is determined by the strength of its dominators in both the
archive and the population. Note that in terms of raw fitness, a low number represents a “good”
solution, i.e. it is not dominated by many members. A zero raw fitness would indicate that the
particular individual is non-dominated. This idea is illustrated in Figure 5.
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
Project Readiness Package
Template Revised Spring 2016
Obj. 2
Obj. 2
A
1/4
A
R=0
S=1
Pareto Front
B
Non-dominated
B
R=0
S=3
3/4
D
R=3
S=2
D
Dominated
7/4
E
7/4
R=5
S=1
C
1/4
F
E
C
F
9/4
R=0
S=1
R=8
S=0
Obj. 1
Obj. 1
Fig. 5. Fitness assignment schemes for SPEA2
The SPEA2 raw fitness assignment scheme utilizes a strength value (S) and a raw fitness value
(R). The strengths of each member are equal to the number of individuals in the population that
they dominate and the raw fitness values are equal to the sum of the strengths of each member’s
dominators. The non-dominated members (A,B,C) have a raw fitness value equal to 0 because
they are not dominated by any other member. Members A and C dominate one member, F, so
their strength is equal to 1. Member B dominates 3 members (D,E,F) so its strength value is 3.
Member D is only dominated by member B, so it is assigned a raw fitness value of 3. It
dominates members E and F, so it is assigned a strength value of 2. This assignment process
continues for each member of the population. Raw fitness is only one part of the SPEA2 fitness
measure. To differentiate between individuals that may, by chance, have identical raw fitness
values, an additional density measure is incorporated. The inverse of the distance to the kth
nearest neighbor is taken to be the density estimate for a given individual. For every member, the
distances to all other members is calculated and sorted in ascending order. The distance to the kth element, denoted, ik , is the point of interest for the density measure. Based on the work of
Silverman the SPEA2 authors determine k as follows:
k  size( P)  size( A)
The density for each individual is then calculated as:
1
D (i )  k
i  2
(13)
(14)
Two is added in the denominator to ensure that D(i) takes a value less than one. Finally, the
fitness of an individual for the SPEA2 algorithm is calculated as:
F (i )  R(i )  D(i)
(15)
The process by which the archive population is updated and maintained is referred to as
environmental selection. For SPEA2, the first step of this process is to copy all non-dominated
individuals from the current archive and population to the next generation’s archive (i.e. At+1).
Non-dominated individuals are easily distinguished, as they are guaranteed to have a fitness
value less than one. This update process is depicted mathematically by:
At 1  i | i  Pt  At  F (i)  1
(16)
At this point, the size of the archive is checked against the pre-defined size limit, denoted N. If
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
Project Readiness Package
Template Revised Spring 2016
size( At 1 )  N
(17)
then the environmental selection process is complete and mating selection can be commenced. If
size( At 1 )  N
(18)
then the archive is filled with the best dominated population members (an easy task given that
the members are already sorted in ascending fitness). The most complicated situation occurs
when
size( At 1 )  N
(19)
When the size of the archive is larger than N, the SPEA2 truncation operator must be utilized.
This algorithm iteratively removes solutions from the archive until it meets the size requirement.
This is accomplished by choosing the individual which has the minimum distance to another
individual at each stage. Ties are broken by the second smallest distance, etc. The goal of this
truncation technique is to maintain the spread of the Pareto front while also preserving the
boundary solutions. The technique is illustrated in Figure 6 and is mathematically depicted in
Equation 20.
i  d j :  0  k  size( At 1 ) :  i k   j k 
 0  k  size( At 1 ) :   0  l  k :  i l   j l    i k   j k 
Obj. 2
A
(20)
1
B
C
3
D
E
2
F
G
Obj. 1
Fig. 6. Illustration of the SPEA2 archive truncation scheme. N = 4 is assumed and the
numbers on the right side indicate the order of elimination.
As Figure 6 depicts, the first solution that is truncated from the Pareto front is the one that is
closest to its nearest neighbor (B). Notice that member A is not removed because it is a boundary
solution. Member G will also not be removed in any truncation scenario. The second elimination
is made as a decision between members E and F. They are the two members with the minimum
distance to each other. Member F is selected to be removed because the distance to its next
nearest neighbor (G) is smaller than member E’s next nearest neighbor (D). The archive size is
set to be four, so one more solution must be removed. The next minimum distance between two
members is between C and D. Member D is chosen to be removed because the distance from DE is smaller than the distance from C-A. The archive is now reduced to its required size of four
members.
† Potential
Concepts:
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
Project Readiness Package
Template Revised Spring 2016
* Project
Deliverables:
Minimum requirements:
 All design documents (e.g., concepts, analysis, detailed drawings/schematics, BOM, test
results)
 working prototype
 technical paper
 poster
 All teams finishing during the spring term are expected to participate in ImagineRIT
Additional required deliverables:
 User’s Manual
† Budget
Information:
Include total budget, any major cost items anticipated, and any special purchasing requirements
from the sponsor(s).

There is NO material cost associated with this project.
* Intellectual
Property:
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
Project Readiness Package
Template Revised Spring 2016
Project Resources
† Required
Resources (besides student staffing):
Describe the resources necessary for successful project completion. When the resource is
secured, the responsible person should initial and date to acknowledge that they have agreed to
provide this support. We assume that all teams with ME/ISE students will have access to the ME
Machine Shop and all teams with EE students will have access to the EE Senior Design Lab, so
it is not necessary to list these. Limit this list to specialized expertise, space, equipment, and
materials.
Faculty list individuals and their area of expertise (people who can provide
specialized knowledge unique to your project, e.g., faculty you will need to consult for
more than a basic technical question during office hours)
Vincent J. Amuso
Steve Boedo
Ferat Sahin
Dorin Patru
Environment (e.g., a specific lab with specialized equipment/facilities, space for very
large or oily/greasy projects, space for projects that generate airborne debris or
hazardous gases, specific electrical requirements such as 3-phase power)
.
Equipment (specific computing, test, measurement, or construction equipment that
the team will need to borrow, e.g., CMM, SEM, )
Desktop & Laptop computerwith Speakers & Microphones.
Materials (materials that will be consumed during the course of the project, e.g., test
samples from customer, specialized raw material for construction, chemicals that must
be purchased and stored)
Initial/
date
Initial/
date
Initial/
date
Initial/
date
Initial/
date
Other
Matlab Software
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
Project Readiness Package
Template Revised Spring 2016
† Anticipated
Staffing By Discipline:
Indicate the requested staffing for each discipline, along with a brief explanation of the
associated activities. “Other” includes students from any department on campus besides those
explicitly listed. For example, we have done projects with students from Industrial Design,
Business, Software Engineering, Civil Engineering Technology, and Information Technology. If
you have recruited students to work on this project (including student-initiated projects),
include their names here.
Dept.
BME
CE
# Req.
0
2
EE
4
ISE
1
ME
Other
0
Expected Activities
Set up Network of PC desktops and/or Laptops
Software interface to collect audio data
Communication Protocol to move data from standalone desktops and/or
laptops to central processing location
(1) Two-way communication between standalone desktops and/or laptops
and central processing location
(1) Phased array system & algorithm development & implementation
(2) Evolutionary Algorithm (EA) development & implementation
Study of office layout and cubicle placement.
Also assist in phased array development & implementation
Also assist in EA development & implementation
* Skills
Checklist:
Indicate the skills or knowledge that will be needed by students working on this project. Please
use the following scale of importance:
1 = must have
2 = helpful, but not essential
3 = either a very small part of the project, or relates to a “bonus” feature
blank = not applicable to this project
Biomedical Engineering
BME Core Knowledge
Matlab
Aseptic lab techniques
Gel electrophoresis
Linear signal analysis and processing
Fluid mechanics
Biomaterials
Labview
Simulation (Simulink)
System physiology
Biosystems process analysis (mass, energy
balance)
Cell culture
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
BME Elective Knowledge
Medical image processing
COMSOL software modeling
Medical visualization software
Biomaterial testing/evaluation
Tissue culture
Advanced microscopy
Microfluidic device fabrication and measurement
Other (specify)
Project Readiness Package
Template Revised Spring 2016
BME Core Knowledge
Computer-based data acquisition
Probability & statistics
Numerical & statistical analysis
Biomechanics
Design of biomedical devices
BME Elective Knowledge
Computer Engineering
CE Core Knowledge
Digital design (including HDL and FPGA)
Software for microcontrollers (including Linux
and Windows)
Device programming (Assembly, C)
Programming: Python, Java, C++
Basic analog design
Scientific computing (including C and Matlab)
Signal processing
Interfacing transducers and actuators to
microcontrollers
CE Elective Knowledge
Networking & network protocols
Wireless networks
Robotics (guidance, navigation, vision, machine
learning, control)
Concurrent and embedded software
Embedded and real-time systems
Digital image processing
Computer vision
Network security
Other (specify)
Electrical Engineering
EE Core Knowledge
Circuit Design (AC/DC converters, regulators,
amplifies, analog filter design, FPGA logic design,
sensor bias/support circuitry)
Power systems: selection, analysis, power budget
System analysis: frequency analysis (Fourier,
Laplace), stability, PID controllers, modulation
schemes, VCO’s & mixers, ADC selection
Circuit build, test, debug (scope, DMM, function
generator
Board layout
Matlab
PSpice
Programming: C, Assembly
Electromagnetics: shielding, interference
EE Elective Knowledge
Digital filter design and implementation
Digital signal processing
Microcontroller selection/application
Wireless: communication protocol, component
selection
Antenna selection (simple design)
Communication system front end design
Algorithm design/simulation
Embedded software design/implementation
Other (specify)
Industrial & Systems Engineering
ISE Core Knowledge
Statistical analysis of data: regression
Materials science
Materials processing, machining lab
Facilities planning: layout, mat’l handling
Production systems design: cycle time, throughput,
assembly line design, manufacturing process
design
Ergonomics: interface of people and equipment
(procedures, training, maintenance)
Math modeling: OR (linear programming,
simulation)
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
ISE Elective Knowledge
Design of Experiment
Systems design – product/process design
Data analysis, data mining
Manufacturing engineering
DFx: manufacturing, assembly, environment,
sustainability
Rapid prototyping
Safety engineering
Project Readiness Package
Template Revised Spring 2016
ISE Core Knowledge
Project management
Engineering economy: Return on Investment
Quality tools: SPC
Production control: scheduling
Shop floor IE: methods, time studies
Computer tools: Excel, Access, AutoCAD
Programming (C++)
ISE Elective Knowledge
Other (specify)
Mechanical Engineering
ME Core Knowledge
3D CAD
Matlab programming
Basic machining
2D stress analysis
2D static/dynamic analysis
Thermodynamics
Fluid dynamics (CV)
LabView
Statistics
Materials selection
RIT – Kate Gleason College of Engineering
Multidisciplinary Senior Design
ME Elective Knowledge
Finite element analysis
Heat transfer
Modeling of electromechanical & fluid systems
Fatigue and static failure criteria
Machine elements
Aerodynamics
Computational fluid dynamics
Biomaterials
Vibrations
IC Engines
GD&T
Linear Controls
Composites
Robotics
Other (specify)
Project Readiness Package
Template Revised Spring 2016