FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
Basics of Optimization & Cohort Intelligence: A Socio-inspired Optimization Technique
1. Basics of Optimization
and
Cohort Intelligence: A Socio-Inspired Optimization
Technique
Anand J Kulkarni
PhD, MASc, BEng, DME
Head & Associate Professor
Department of Mechanical Engineering
Symbiosis Institute of Technology
Symbiosis International University
Pune 412 115, MH, India
Email: anand.kulkarni@sitpune.edu.in;
kulk0003@ntu.edu.sg
Ph: 91 20 3911 6468
Odette School of Business
University of Windsor
401 Sunset Avenue
Windsor, Ontario N9B 3P4
Canada
E-mail: kulk0003@uwindsor.ca
Ph: 1 519 253 3000 (x4939)
2. Agenda
• Basics of Optimization
• Contemporary Algorithms
• Cohort Intelligence
• Validation
• Test on Combinatorial Problems
• Applications to Real World Problems
• Recent Developments and Future Directions
10/25/2017 Symbiosis International University & University of Windsor 2
3. What is optimization about?
• Extreme states (i.e. minimum and maximum states out of many or possibly
infinitely many)
Ex. Natural (physical) stable equilibrium state is generally a ‘minimum potential
energy’ state.
• Human activities: to do the best in some sense (Intrinsic Human Nature)
• set a record in a race (shortest/minimum time, etc.)
• retail business (maximize the profit, etc.)
• construction projects (minimize cost, time, etc.)
• power generator design (maximize efficiency, minimize weight, etc.)
• Best job out of several choices
310/25/2017 Symbiosis International University & University of Windsor
4. What is optimization about?
• Understand and analyze the natural/physical phenomena
• Mathematically model it
• Solve real world problems
410/25/2017 Symbiosis International University & University of Windsor
5. What is optimization about?
• Real world issues:
• Requirements and constraints imposed on products, systems,
processes, etc.
• Creating feasible design (solution)
• Creating a best possible design (solution)
• “Design optimization”: highly complex, conflicting constraints and
considerations, etc.
510/25/2017 Symbiosis International University & University of Windsor
6. Importance of Optimization
• Greater concern about the limited energy, material, economic
sources, etc.
• Heightened environmental/ecological considerations
• Increased technological and market competition
610/25/2017 Symbiosis International University & University of Windsor
7. A Simple Example
• 5 X 7 metal sheet
• 𝑥 can take different values between 0 and 2.5
• Infinite box designs (solutions)
• Aim: Biggest box volume (Maximization)
710/25/2017 Symbiosis International University & University of Windsor
8. A Simple Example
• Setting obtain stationary points
or
'
0f x
0.96x 3.04x
0.96 15.02 ; 3.04 3.02f f
3 2
' 2
2
''
2
5 2 7 2 4 24 35 , 0 2.5
12 48 35
24 48
f x x x x x x x x
df
f x x x
dx
d f
f x x
dx
810/25/2017 Symbiosis International University & University of Windsor
10. A Simple Example
• Roots and Optima
• Find roots by setting
• For maximum and
• For minimum and
0f x
'
0f x
''
0f x
''
0f x
'
0f x
1010/25/2017 Symbiosis International University & University of Windsor
14. Local and Global Optimum
• An objective function is at its local minimum at the point if
for all feasible within its small neighborhood of
14
*
f fX X
*
Xf
X
*
X
10/25/2017 Symbiosis International University & University of Windsor
15. Local and Global Optimum
• An objective is at its global minimum at the point if
for all feasible .
15
*
f fX X*
Xf
X
10/25/2017 Symbiosis International University & University of Windsor
16. Design of a Can
1610/25/2017 Symbiosis International University & University of Windsor
18. Basic Definitions
• A design problem is characterized by a set of design
variables (decision variables)
• Single Variable
• Multi-Variable
where
2
min 2 logf x x x
5
1 2 1 2
5
1 2
min , 2 log
2 log
f x x x x
f x x
X
1 2,x xX
1810/25/2017 Symbiosis International University & University of Windsor
19. • 3D view of 2D optimization
1910/25/2017 Symbiosis International University & University of Windsor
20. Basic Definitions
• Design variables
• Continuous (any value between a specific interval)
• Discrete (the value from a set of distinct numerical values)
• Ex. Integer values, (1, 4.2, 6 11, 12.9, 25.007), binary (0, 1), etc.
• Combinatorial Optimization Problem
• Mixed (discrete & continuous) variables
2010/25/2017 Symbiosis International University & University of Windsor
24. Basic Definitions
• Unconstrained Optimization Problems
• No restrictions (Constraints) imposed on design variables
• Constrained Optimization Problems
• Restrictions (constraints) are imposed on design variables and the
final solution should satisfy these constraints, i.e. the final solution
should at least be feasible.
• The ‘best’ solution comes further
2410/25/2017 Symbiosis International University & University of Windsor
25. Basic Definitions
• Depending on physical nature of the problem: ‘Optimal Control
Problem’
• Decomposing the complex problem into a number of simpler sub-
problems
• Linear Programming (LP): If the objective and constraints are linear
• Non-Linear programming (NLP): If any of it is non-linear
2510/25/2017 Symbiosis International University & University of Windsor
26. General Problem Statement
• Side constraints
• (Splitting into two inequality constraints)
:
0 , 1,...,
0 , 1,...,
, 1,...,
j
k
i
Minimize f
Subject to
g j m
h k l
x i n
X
X
X
l u
i i ix x x
2610/25/2017 Symbiosis International University & University of Windsor
27. Active/Inactive/Violated Constraints
• Inequality Constraints
27
1
2
3
4
,
12 5 3000
10 14 4000
50 50
50 50
f f B R
g B R
g B R
g B B
g R R
X
10/25/2017 Symbiosis International University & University of Windsor
28. Active/Inactive/Violated Constraints
• The set of points at which an inequality constraint is active forms a
constraint boundary which separates the feasible region points from
the infeasible region points.
2810/25/2017 Symbiosis International University & University of Windsor
29. Active/Inactive/Violated Constraints
• An inequality constraint is said to be violated at a point , if it is
not satisfied there i.e. .
• If is strictly satisfied i.e. . Then it is said to be inactive at
the point .
• If is satisfied at equality i.e. . Then it is said to be active
at the point .
29
jg
0jg X
jg 0jg X
X
X
0jg Xjg
X
10/25/2017 Symbiosis International University & University of Windsor
31. Active/Inactive/Violated Constraints
• Based on these concepts, equality constraints can only be either
active i.e. or violated i.e. at any point .
31
0jh X 0jh X X
10/25/2017 Symbiosis International University & University of Windsor
32. Active/Inactive/Violated Constraints
• Equality & inequality constraints
32
1 2
1 1 2
2 1 2
3 1 2
1 1 1
2 2 2
,
4 2 12
1
2 4
0 0
0 0
f f x x
h x x
h x x
h x x
g x x
g x x
X
Design Space
Feasible Region
Infeasible Region
10/25/2017 Symbiosis International University & University of Windsor
33. Convexity
• A set of points is a convex set, if for any two points in the set, the
entire straight line segment joining these two points is also in the set.
3310/25/2017 Symbiosis International University & University of Windsor
34. Convexity
• A function f(X) is convex if it is defined over a convex set and for any
two points of the graph f(X), the straight line segment joining these
two points lies entirely above or on the graph.
3410/25/2017 Symbiosis International University & University of Windsor
35. Algorithms
• Exact Methods/Algorithms
• Approximations Methods/Algorithms
• Artificially Intelligent methods
• Bio-/Nature-inspired Methods
• Self-organizing Systems
10/25/2017 Symbiosis International University & University of Windsor 35
40. Cohort Intelligence
• A Socio-inspired Self Organizing System
• Includes inherent, self realized and rational learning
• Self control and ability avoid obstacles (jumps out of ditches/local solutions)
• Inherent ability to handle constraints
• Inherent ability of handling uncertainty
10/25/2017 Symbiosis International University & University of Windsor 40
41. Conference Publications
• Kulkarni, A.J., Durugkar I.P., Kumar M. (2013): “Cohort Intelligence: A
Self Supervised Learning Behavior”, in Proceedings of IEEE
International Conference on Systems, Man and Cybernetics,
Manchester, UK, 13-16 October 2013, pp. 1396-1400
• Kulkarni, A.J., Baki, F., Chaouch, B. (2014): A New Variant of the
Assignment Problem: Application, NP-hardness and Algorithms,
Optimization Days, Montreal, Canada, May 5-7, 2014
10/25/2017 Symbiosis International University & University of Windsor 41
42. Book Chapters
• Gaikwad, S., Joshi, R., Kulkarni, A.J. (2016): “Cohort Intelligence and
Genetic Algorithm along with Modified Analytical Hierarchy Process to
Recommend an Ice Cream to a Diabetic Patient”, Advances in Intelligent
and Soft Computing, Satapathy, Bhateja, Joshi (Eds), Springer, Vol 468, pp.
279-288
• Gaikwad, S., Joshi, R., Kulkarni, A.J. (2015): “Cohort Intelligence and
Genetic Algorithm along with AHP to recommend an Ice Cream to a
Diabetic Patient”, Lecture Notes in Computer Science, Vol. 9873, Springer,
pp. 40-49
• Shastri, A.S., Jadhav, P.S., Kulkarni, A.J., Abraham, A. (2016): “Solution to
Constrained Test Problems using Cohort Intelligence Algorithm”, Advances
in Intelligent and Soft Computing, Vol. 424, Springer, pp. 427-435
10/25/2017 Symbiosis International University & University of Windsor 42
43. Journal Publications
• Sarmah, D., Kulkarni, A.J. (2017): “Image Steganography Capacity Improvement using
Cohort Intelligence and Modified Multi Random Start Local Search Methods Arabian
Journal for Science and Engineering”, (In Press: Arabian Journal for Science and
Engineering)
• Patankar, N.S., Kulkarni, A.J. (2017): "Variations of Cohort Intelligence", (In Press: Soft
Computing)
• Kale I.R., Kulkarni, A.J. (2017): “Cohort Intelligence Algorithm for Discrete and Mixed
Variable Engineering Problems”, (In Press: International Journal of Parallel, Emergent and
Distributed Systems)
• Shah, P., Agashe, S., Kulkarni, A.J. (2017): “Design of Fractional PID Controller using
Cohort Intelligence Method”, (In Press: Frontiers of Information Technology & Electronic
Engineering)
• Dhavle S.V., Kulkarni, A.J., Shastri A., Kale I.R. (2017): “Design and Economic Optimization
of Shell-and-Tube Heat Exchanger using Cohort Intelligence Algorithm” (In Press: Neural
Computing and Applications)
10/25/2017 Symbiosis International University & University of Windsor 43
44. Journal Publications
• Kulkarni, O., Kulkarni, N., Kulkarni, A.J., Kakandikar, G. (2016): “Constrained
Cohort Intelligence using Static and Dynamic Penalty Function Approach for
Mechanical Components Design” (In Press: International Journal of Parallel,
Emergent and Distributed Systems)
• Kulkarni, A.J., Shabir, H. (2016): “Solving 0-1 Knapsack Problem using Cohort
Intelligence Algorithm”. International Journal of Machine Learning and
Cybernetics, 7(3), pp. 427-441
• Kulkarni, A.J., Baki, M.F., “Chaouch, B.A. (2016): “Application of the Cohort-
Intelligence Optimization Method to Three Selected Combinatorial Optimization
Problems”, European Journal of Operational Research, 250(2), pp. 427-447
• Krishnasamy, G., Kulkarni A.J., Paramesaran, R. (2014): “A hybrid approach for
data clustering based on modified cohort intelligence and K-means”, Expert
Systems with Applications, 41(13), pp. 6009-6016
10/25/2017 Symbiosis International University & University of Windsor 44
45. Book
• Kulkarni, A.J., Krishnasamy, G., Abraham,
A.: “Cohort Intelligence: A Socio-inspired
Optimization Method”, Intelligent Systems
Reference Library, 114 (2017) Springer,
(DOI 10.1007/978-3-319-44254-9), (ISBN:
978-3-319-44254-9)
10/25/2017 Symbiosis International University & University of Windsor 45
46. What is a Cohort
• A group of candidates interacting and competing with one another to
achieve some individual goal which is inherently common to all the
candidates.
Exhibits a Self Organizing System
10/25/2017 Symbiosis International University & University of Windsor 46
47. What is a Cohort?
• They (We??) need a supervisor like a friend/colleague which can work
with us, right?
10/25/2017 Symbiosis International University & University of Windsor 47
48. Can Individuals Learn from Peers?
• “Hole in the Wall” experiment by Dr. Sugata Mitra (1999)
• With no supervision or formal teaching, children can teach
themselves and each other, if motivated by curiosity and peer
interest.
http://www.hole-in-the-wall.com/MIE.html10/25/2017 Symbiosis International University & University of Windsor 48
49. Cohort Intelligence Algorithm
• Initialize number of candidates in the cohort, quality variations ,
and set up interval reduction factor
• Step 1 The probability associated with the behavior being followed by
every candidate in the cohort is calculated
• Step 2 Using roulette wheel approach every candidate selects
behavior to follow from within the available choices
C t
r
10/25/2017 Symbiosis International University & University of Windsor 49
50. Cohort Intelligence Algorithm
• Step 3 Every candidate shrinks/expands the sampling
interval of every quality based on whether condition
of saturation is satisfied
• Step 4 Every candidate forms behaviors by sampling
the qualities from within the updated sampling
intervals
• Step 5 Every candidate follows the best behavior from
within its behaviors
• Step 6 Cohort behavior saturated?
• NO? go to Step 1
10/25/2017 Symbiosis International University & University of Windsor 50
51. Cohort Intelligence Algorithm
• Step 7 Convergence?
• NO? go to Step 1
• Accept the current cohort behavior as final solution
10/25/2017 Symbiosis International University & University of Windsor 51
52. Cohort behavior
saturated?
Y
N
START
Initialize number of candidates in the cohort,quality
variations, and set up interval reduction factor
STOP
Accept the current
cohort behavioras
final solution
Every candidateshrinks/expandsthe sampling interval of
every quality based on whether condition of saturationis
satisfied
Using roulette wheel approach every candidate selects
behavior to follow from within the available choices
The probability associated with the behavior being
followed by every candidatein the cohort is calculated
N
Every candidateforms behaviorsby sampling the qualities
from within the updated sampling intervals
Every candidatefollows the best behaviorfrom within its
behaviors
Convergence?
Y
1
1
, 1,...,
1
c
c
C
c
c
f
p c C
f
x
x
? ? ?
2 , 2
c c c
i i i i ix r x r
, 1,...,c
f c Cx
1
,..., ,...,C c C
f f fF x x x
1Minimize ,... ,...,
Subject to , 1,...,
i N
lower upper
i i i
f f x x x
x i N
x
1,...,c C 1 ,... ,...,c c c c
i Nx x xx c
f x
Candidates Qualities Behavior
Possibility of
being followed
Neighborhood
space
New Qualities &
Behaviors
Cohort
Solutions
10/25/2017 Symbiosis International University & University of Windsor 52
57. Problem
RHPSO [13] CPSO [16]
LDWPSO
[16]
SQP [15] Proposed CI
Best
Mean
Worst
Best
Mean
Worst
Best
Mean
Worst
Best
Mean
Worst
Best
Mean
Worst
FE SD Time (sec)
Sphere
1.5000E-323
3.5078E-245
5.0380E-248
1.4356E-81
3.4213E-12
1.7103E-10
1.5387E-06
1.2102E-04
1.1486E-03
3.5657E-28
2.5749E-27
8.8173E-27
2.0000E-15
2.4900E-06
1.7780E-05
5 0.80
18750 4.5800E-03 1.55
Rosenbrock
1.5606E-08
1.2061E-07
3.0398E-07
1.1856E-08
9.3949E-03
9.0066E-02
2.8453E_03
3.1101E+00
1.1050E+01
7.5595E-12
1.4352E+00
3.9866E+00
0.0000E+00
0.0000E+00
0.0000E+00
5 0.80
9750 0.0000E+00 5.20
Ackley
0.0000E+00
0.0000E+00
0.0000E+00
8.8178E-16
1.5952E-08
6.3330E-07
1.3078E-04
5.9934E-03
2.5325E-02
1.5245E+01
1.9090E+01
1.9959E+01
1.2322E-07
2.0911E-07
2.6499E-07
5 0.85
11250 4.3200E-08 1.50
Griewank
0.0000E+00
0.0000E+00
0.0000E+00
0.0000E+00
2.1287E-10
6.4174E-09
1.6949E-02
1.7072E-01
7.2835E-01
2.8879E-09
3.5357E-01
3.6312E+00
7.3960E-03
1.7100E-02
4.9183E-02
5 0.997
18750 8.8300E-03 2.00
0 20 40 60 80 100 120 140 160
0
200
400
600
800
1000
1200
Learning Attempts
Behavior
Candidate 1
Candidate 2
Candidate 3
Candidate 4
Candidate 5
0 50 100 150 200 250
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Learning Attempts
Behavior
Candidate 1
Candidate 2
Candidate 3
Candidate 4
Candidate 5
C r
10/25/2017 Symbiosis International University & University of Windsor 57
58. Cohort Intelligence
• Combinatorial Problems
• Packing Problem
10/25/2017 58Symbiosis International University & University of Windsor
1
1
where (
( )
0,, ,) 1 1
N
i i
i
N
i i
i
i
Maximize f
Subject to f W
x if w x N
v x
v
w
w
60. Feasibility-based Rule
1. If the solution of candidate c is feasible:
1.1. Adds a randomly chosen object from the candidate being followed
respecting feasibility.
1.2. Replaces a randomly chosen object with another randomly chosen one
from the candidate being followed respecting feasibility.
2. If the candidate c is infeasible:
• 2.1. Removes a randomly chosen object from within its knapsack.
• 2.2. Replaces a randomly chosen object with another randomly chosen one
from the candidate being followed, such that the total weight of the
candidate c decreases.
10/25/2017 Symbiosis International University & University of Windsor 60
61. Candidate 1 follow 3
10/25/2017 Symbiosis International University & University of Windsor 61
69. Cost Comparison for Case 1
51507 49259 46453 44559 44536 44301.66 44259.01 44132.519
12973
5818
6778.2
6233.8 6046 5924.373 5914.058 5873.6607
0
10000
20000
30000
40000
50000
60000
70000
Original
Study
GA
PSO
ABC
BBO
ITHS
I-ITHS
CI
Total discounted operating cost
Capital investment
6910/25/2017 Symbiosis International University & University of Windsor
73. Cost Comparison for Case 3
16549
19163 18614 17893 18059 18273 18209 18447.6373
27440
1671 1696 1584.2 1251.5 1419 1464 2356.2566
0
5000
10000
15000
20000
25000
30000
35000
40000
45000
50000
OriginalStudy
GA
PSO
ABC
BBO
ITHS
I-ITHS
CI
TotalCost(€)
Total discounted operating cost
Capital investment
7310/25/2017 Symbiosis International University & University of Windsor
74. Performance Details of CI
74
Case studies
Solutions Standard
Deviation
Avg. No. of
Function
Evaluations (FE)
Avg. Comp.
Time (sec)Best
Mean
Worst
Case 1
50006.17969
0.2675 2190 0.2474
50006.52636
50006.93099
Case 2
19298.18004
0.4065 2183 0.3698
19298.65499
19299.50914
Case 3
20803.89398
0.1252 1998 0.2471
20804.06749
20804.31439
10/25/2017 Symbiosis International University & University of Windsor
75. Closeness of CI Solution with Other Algorithms
75
Case studies
Total Cost (€) of
CI solution
Referred
algorithms
Solutions of Total
Cost (€)
Closeness to the
Best Reported
Solution (%)
Case 1
Original Study 64480 22.45 ↑
GA 55077 9.21 ↑
PSO 53231.1 6.06 ↑
50006.1797 ABC 50793 1.55 ↑
BBO 50582 1.14 ↑
ITHS 50226 0.44 ↑
I-ITHS 50173 0.33 ↑
Case 2
Original Study 27020 28.58 ↑
GA 20303 4.95 ↑
PSO 19922.6 3.13 ↑
19298.1800 ABC 20225 4.58 ↑
BBO 19810 2.58 ↑
ITHS 20211 4.52 ↑
I-ITHS 20182 4.38 ↑
Case 3
Original Study 43989 52.70 ↑
GA 20834 0.02 ↑
PSO 20310 2.43 ↓
20803.8940 ABC 19478 6.81 ↓
BBO 19310 7.74 ↓
ITHS 19693 5.64 ↓
I-ITHS 19674 5.74 ↓10/25/2017 Symbiosis International University & University of Windsor
76. Real World Combinatorial Problems
Anand J Kulkarni
Odette School of Business,
University of Windsor,
401 Sunset Avenue,
Windsor, Ontario N9B 3P4,
Canada
E-mail: kulk0003@uwindsor.ca
Ph: 1 519 253 3000 (x4939)
M F Baki
Odette School of Business,
University of Windsor,
401 Sunset Avenue,
Windsor, Ontario N9B 3P4,
Canada
E-mail: fbaki@uwindsor.ca
Ph: 1 519 253 3000 (x3118)
Ben A Chaouch
Odette School of Business,
University of Windsor,
401 Sunset Avenue,
Windsor, Ontario N9B 3P4,
Canada
E-mail: chaouch@uwindsor.ca
Ph: 1 519 253-4232 (x3149)
10/25/2017 Symbiosis International University & University of Windsor 76
77. New Variant of the Assignment Problem
Row Circular Matrix 𝐶 = 𝐶𝑖𝑗
𝐶 =
1 2 3 4
5 6 7 8
9 10 11 12
13 14 15 16
1 2 3 4
7 8 5 6
12 9 10 11
14 15 16 13
1
3
2
4
• Find a permutation that minimizes the maximum column sum
of the rotated matrix.
• It is a variant of the assignment problem equivalent to finding
a permutation that minimizes the minimum column sum of the
rotated matrix.
• 3-Partition problem reduced to the new variant of assignment
problem proving its strong NP-hardness.
10/25/2017 Symbiosis International University & University of Windsor 77
80. We refer this problem as
Cyclic Bottleneck Assignment Problem (CBAP).
Cyclic: Row circularity of matrix 𝐶
Bottleneck: minmax objective
Assignment: problem’s close affinity to the classical assignment
problem σ𝑖=1
𝑛
𝐶𝑖,𝜋 𝑖
10/25/2017 Symbiosis International University & University of Windsor 80
81. Applications
40 30 22 13
36 29 23 7
32 30 22 7
36 32 23 17
𝑴 𝑻 𝑾 𝑹
𝑫 𝟏
𝑫 𝟐
𝑫 𝟑
𝑫 𝟒
𝐼 𝑘 = 144 121 90 44
𝑍 𝐶∗𝜋1
= 105
22 13 40 30
29 23 7 36
7 32 30 22
36 32 23 17
𝑴 𝑻 𝑾 𝑹
𝑫 𝟏
𝑫 𝟐
𝑫 𝟑
𝑫 𝟒
𝐼 𝑘 = 94 100 100 105
𝜋∗1 = 3, 4, 2, 1
• Healthcare
• The problem arises in minimizing congestion in the recovery unit
• Planning horizon of n days with cyclic scheduling
• Keep the maximum number of patients as low as possible to reduce
the requirement of beds, nurses and other variable costs
Beds, Nurses, Variable
Costs, etc.
10/25/2017 Symbiosis International University & University of Windsor 81
82. Applications
40 30 22 13
36 29 23 7
32 30 22 7
36 32 23 17
𝑴 𝑻 𝑾 𝑹
𝑺 𝟏
𝑺 𝟐
𝑺 𝟑
𝑺 𝟒
𝐼 𝑘 = 144 121 90 44
𝑍 𝐶∗𝜋1
= 105
22 13 40 30
29 23 7 36
7 32 30 22
36 32 23 17
𝑴 𝑻 𝑾 𝑹
𝑺 𝟏
𝑺 𝟐
𝑺 𝟑
𝑺 𝟒
𝐼 𝑘 = 94 100 100 105
𝜋∗1 = 3, 4, 2, 1
• Inventory Management
• Minimizing the maximum space requirement in a retail store
• Planning horizon of n days
• Suppliers follow a cyclic schedule
Space Requirement
10/25/2017 Symbiosis International University & University of Windsor 82
84. 𝑡 = 1
The element from 𝜋∗3 = 1, 4, 3, 2 being followed: 3
Its location in the permutation 𝜋∗1 = 3, 4, 2, 1 : 1,1
The updated permutation 𝜋1,1: 2, 4, 3, 1
Updated circular matrix 𝐶∗1 =
22 13 40 30
29 23 7 36
7 32 30 22
36 32 23 17
30 22 13 40
23 7 36 29
30 22 7 32
36 32 23 17
Maximum column sum 𝑍 𝐶1 1: 119
𝑡 = 2 The element from 𝜋3 = 1, 4, 3, 2 being followed: 1
Its location in the permutation 𝜋1 = 3, 4, 2, 1 : 1,4
The updated permutation 𝜋1,2: 1, 4, 2, 3
Updated circular matrix 𝐶∗1 =
22 13 40 30
29 23 7 36
7 32 30 22
36 32 23 17
22 13 40 30
23 7 36 29
22 7 32 30
23 17 36 32
Maximum column sum 𝑍 𝐶1 2: 144
𝑍 𝐶∗𝜋1
= min 𝑍 𝐶 𝜋1 1
, 𝑍 𝐶 𝜋1 2
and
associated permutation 𝜋∗1:
119 and
𝟐, 𝟒, 𝟑, 𝟏
90 44 144 121
119 83 79 118
10/25/2017 Symbiosis International University & University of Windsor 84
85. Numerical Experiments and Results
0 50 100 150 200 250 300 350 400
4660
4680
4700
4720
4740
4760
4780
4800
4820
4840
Learning Attempts
CICandidateSolutions
CI Parameters Cases
Candidates Variations
25 5
Every case - 10 instances & solved 20 times
10/25/2017 Symbiosis International University & University of Windsor 85
86. Numerical Experiments and Results
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
5x5 6x6 7x7 8x8 9x9 10x10 11x11 12x12 13x13
Avg.%gapofLBwithIP
n n
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
Avg%gapofCIwithLB
n n
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
5x5 6x6 7x7 8x8 9x9 10x10 11x11 12x12 13x13
Avg.%gapofCIwithIP
n n
0
200
400
600
800
1000
1200
5x5 6x6 7x7 8x8 9x9 10x10 11x11 12x12 13x13
IP:AverageCPUTime
n n
10/25/2017 Symbiosis International University & University of Windsor 86
87. Numerical Experiments and Results
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
5x5 6x6 7x7 8x8 9x9 10x10 11x11 12x12 13x13
CIAverageFE
n n
0
10000
20000
30000
40000
50000
60000
70000
80000
90000
15x15 20x20 25x25 30x30 35x35 40x40 45x45 50x50
CIAverageFE
n n
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
5x5 6x6 7x7 8x8 9x9 10x10 11x11 12x12 13x13
CIAverageCPUTime
n n
0
5
10
15
20
25
30
15x15 20x20 25x25 30x30 35x35 40x40 45x45 50x50
CIAverageCPUTime
n n
10/25/2017 Symbiosis International University & University of Windsor 87
88. Numerical Experiments and Results
Problem Size Average
% gap
LP vs. CPLEX
CPLEX:
Average CPU
Time
(Sec)
CI Method
Average
% gap
LP vs. CI
Average
% gap
CPLEX vs. CI
Standard
Deviation
(SD)
CPU Time
(Sec)
Function
Evaluations
(FE)
5 4.3165 0.27 4.5496 0 0 0.038084 1620
6 3.3520 0.29 3.4857 0 0 0.082466 3169
7 2.4426 0.41 2.5067 0 0 0.141646 5070
8 1.7478 0.73 1.7815 0 0.659745 0.206037 6988
9 1.3584 4.10 1.3778 0 1.920292 0.301112 9136
10 0.9259 13.80 0.9473 0.0120 3.265155 0.359756 10750
11 0.8597 98.31 0.9739 0.1057 2.961605 0.443356 12190
12 0.7083 260.59 0.9296 0.2146 3.045473 0.537049 13601
13 0.5632 1072.93 0.8675 0.2993 2.695255 0.663312 15477
15 -- -- 0.8040 -- 2.832436 0.831481 17827
20 -- -- 0.6492 -- 2.727449 1.659961 26085
25 -- -- 0.5431 -- 2.636055 3.138706 35435
30 -- -- 0.4502 -- 2.668305 5.005378 43715
35 -- -- 0.4031 -- 2.586950 7.842689 51145
40 -- -- 0.3417 -- 2.696257 11.972532 61610
45 -- -- 0.3130 -- 2.468574 17.019936 71835
50 -- -- 0.2784 -- 2.545943 23.980858 81207
10/25/2017 Symbiosis International University & University of Windsor 88
89. Sea Cargo Mix Problem
Deciding a sea cargo shipping schedule for freight bookings accepted in multi-
period planning horizon (Ang et al. 2007)
Maximize: profit generated by all freight bookings
Subject to:
• demand for empty containers at the port of origin is less than or equal to
the number of available empty containers at the port of origin in each
period
• total weight and volume of cargoes which will be carried to a destination
port in a period is less than or equal to the total available weight and
volume capacity of shipment to that port in a that period.
• each cargo may be carried in a certain period on or before its due date or
be refused to carry in the time horizon
• each cargo can be either accepted at its total quantity or be turned down
10/25/2017 Symbiosis International University & University of Windsor 89
90. Total Volume and Weight
Capacity
Volume and Weight Capacity to
different destinations in
particular period
… …
… …
Destination
Ports
Origin Port
Cargoes:
Weight,
Volume, Due
date, port of
destination
Maximize:
Profit
10/25/2017 Symbiosis International University & University of Windsor 90
95. Selection of Cross-Border Shippers Problem
• NAFTA
• Increase in Traffic between USA-Canada-Mexico
• Cross-Border Compliance
• Avoidance of Delays at Check Points
• Reduce Transportation Time, Cost, etc.
10/25/2017 Symbiosis International University & University of Windsor 95
96. Selection of Cross-Border Shippers Problem
• Goals:
• ‘volume capacity’: total volume of containers assigned to a
shipper does not exceed its maximum capacity
• ‘fund availability’: ensures that the total expenditure
should not exceed the available fund allotted for a
particular period
• ‘due date delivery’: processing time for a good should not
exceed the due delivery date
• ‘number of maximum allowable non-compliant shippers’
• Type of good and handling ability of Shipper
10/25/2017 Symbiosis International University & University of Windsor 96
97. Shippers:
Individual Volume and
Weight Capacity,
ability to handle type
of good,
Cross-Border
Compliant/non-
compliant,
fixed/variable cost of
shipping
… …
Containers:
Weight,
Volume, Due
date, type of
good
Goals:
Fund, Due date,
maximum
allowable non-
compliant shippers
Selection of Cross-Border Shippers Problem
10/25/2017 Symbiosis International University & University of Windsor 97
98. Selection of Cross-Border Shippers Problem
Single Period Multi periodMathematical Formulation
10/25/2017 Symbiosis International University & University of Windsor 98
100. Path Planning and Obstacle Avoidance for
Swarm Robots using Cohort Intelligence
Palash Roychowdhury
Prakhar Shrivastava
Rishi Devarakonda
Siddarth Mehra
Siddharth Basu
Dr. Anand J Kulkarni
10/25/2017 Symbiosis International University & University of Windsor 100
101. Problem Statement/Objectives
• The application is relevant to search and rescue in alien territory as
well as establishment
10/25/2017 Symbiosis International University & University of Windsor 101
102. Problem Statement/Objectives
• Every robot is assumed
to have two sensors:
• light sensor
• proximity sensor
10/25/2017 Symbiosis International University & University of Windsor 102
𝐿1
𝐿2
𝐿3
𝐿4
𝐵1
𝐵2
𝐵3
𝐵4
𝐿
Arena
Every robot is represented by
𝐵𝑖 , 𝑖 = 1,2, … , 𝑁
Receives light with an intensity
𝐿𝑖 , 𝑖 = 1,2, … , 𝑁
The goal/objective is to collectively reach the light source 𝐿 which is a possible
exit door of the arena.
103. Problem Statement/Objectives
• Path Planning of
swarm robots and
avoidance of
obstacles in complex
in arena using:
Linear Probability
Approach
Exponential
Probability Approach
10/25/2017 Symbiosis International University & University of Windsor 103
104. Roulette Wheel Probability Distribution
10/25/2017 Symbiosis International University & University of Windsor 104
POINT ROULETTE WHEEL POSITION PROBABILITY
START POINT END POINT
𝐵1 0 0.40 0.40
𝐵2 0.40 0.55 0.15
𝐵3 0.55 0.75 0.20
𝐵4 0.75 1.00 0.25
𝑀𝑎𝑥𝑖𝑚𝑖𝑧𝑒
𝑖=1
𝑁
𝐿𝑖
𝑃𝑖= ൗ
𝐿 𝑖
𝑧
σ𝑖=1
𝑁 𝐿 𝑖
𝑧
𝑃𝑖 = (𝑃1, 𝑃2, 𝑃3,………, 𝑃𝑖)
σ𝑖=1
𝑁
𝑃𝑖 = 1
105. Motion of Robots
Random Number
Generation
Condition 1: Number
lies in its own region
Condition 2: Number
lies in other robot’s
region
10/25/2017 Symbiosis International University & University of Windsor 105
𝑝1 + 𝑝2 + 𝑝3 +𝑝4 +𝑝5= 1
Roulette Wheel
Selection
𝑟𝑎𝑛𝑑 = 0,1
Learning Attempt
𝑛
𝐵2
𝐵1
𝐵3
𝐵4
𝐵5
𝑝1 𝑝1
+ 𝑝2
𝑝1
+ 𝑝2
+ 𝑝3 𝑝1 + 𝑝2 + 𝑝3 +𝑝4
0
𝑥
𝑦
Learning Attempt
𝑛 + 1
𝐿1
𝐿2
𝐿3
𝐿4
𝐿5
𝐵2
𝐵1
𝐵3
𝐵4
𝐵5
𝑥
𝑦
𝐿1
′
𝐿2
′
𝐿4
′
𝐿5
′
𝐵1
𝐵2
𝐵5
𝐵4
𝐿3
′
𝐵3
Following itself Following other candidate
106. 4 Independent Cases, 20 Independent Runs
10/25/2017 Symbiosis International University & University of Windsor 106
Linear Probability
𝑃𝑖 = ൗ
𝐿 𝑖
σ 𝑖=1
𝑁 𝐿 𝑖
No Obstacle
Case (NOC)
Rectangular
Obstacles Case
(ROC)
Multiple Rectangular
Obstacles Case
(MROC)
Cluttered Polygons
Obstacles Case
(CPOC)Exponential Probability
𝑃𝑖 = ൗ
𝐿 𝑖
𝑍
σ 𝑖=1
𝑁 𝐿 𝑖
𝑍
𝟓 (five) robots,
CI algorithm was coded in
MATLAB R2014b on
Windows 8.1,
Intel Core i7 2.3GHz processor speed
and
8GB RAM.
107. No Obstacle Case (NOC): Linear Probability &
Randomly Initiated Robots
10/25/2017 Symbiosis International University & University of Windsor 107
114. 114
Illustration 1 Illustration 2
Cluttered Polygonal Obstacle Case (CPOC): Exponential
Probability & Randomly Initiated Robots
115. Performance of CI with Linear and Exponential Probability for No Obstacle Case (NOC)
CI with Linear Probability CI with Exponential Probability
Config.
Mean
(FE)
SD
(FE)
Mean
Time
(Seconds)
SD Time
(Seconds)
Mean
Distance
Travelled
Mean
Total Time
(Seconds)
SD Total
Time
(Seconds)
Mean
(FE)
SD
(FE)
Mean
Time
(Seconds)
SD Time
(Seconds)
Mean
Distance
Travelled
Mean
Total Time
(Seconds)
SD Total
Time
(Seconds)
1 1067 442.95 1.30 0.80 4.02 2.03 0.91 852 370.47 1.43 0.80 3.04 1.70 0.80
2 1345 423.19 2.58 0.85 5.75 2.84 1.14 1145 192.88 2.37 0.60 4.61 2.39 0.61
3 1087 517.33 2.04 0.72 3.32 2.81 0.84 1032 411.36 2.26 0.86 3.05 2.43 0.72
4 1240 460.78 2.02 0.67 6.00 2.21 0.84 1095 217.65 2.18 0.53 5.27 2.21 0.49
5 987 425.51 2.31 0.60 4.68 2.56 0.64 947 347.80 2.04 0.59 4.19 2.17 0.64
6 1347 439.04 2.27 0.85 4.23 2.58 0.92 852 367.35 1.72 0.74 3.54 1.80 0.76
7 1115 982.74 2.69 2.38 3.98 2.76 3.09 1010 539.61 2.33 0.99 3.38 2.40 1.29
8 855 359.71 1.44 0.59 5.07 1.63 0.72 742 168.11 1.46 0.54 4.23 1.49 0.52
9 975 342.85 1.82 0.87 4.13 1.88 0.92 935 214.44 1.78 0.60 3.79 1.81 0.59
10 1192 341.24 2.91 0.69 4.77 3.26 0.72 1140 133.74 2.58 0.63 4.08 2.63 0.56
11 1505 514.46 1.54 2.40 4.17 2.48 2.60 1070 314.58 1.60 0.77 3.20 1.85 0.65
12 1287 2238.89 2.25 0.72 7.60 2.65 5.74 1180 173.49 2.46 0.60 4.17 2.49 0.57
13 1230 237.86 2.36 0.60 4.93 2.62 0.64 1022 239.65 2.28 0.52 4.19 2.41 0.57
14 1170 856.73 1.90 0.73 5.20 2.23 1.34 1092 314.05 1.77 0.74 3.96 1.85 0.70
15 1382 332.99 2.23 1.32 5.10 2.25 1.23 1147 305.99 2.07 0.93 4.06 2.08 0.93
16 1832 915.33 2.89 0.68 5.78 3.43 2.17 1147 183.23 2.46 0.55 3.49 2.49 0.55
17 1152 1051.20 2.15 1.04 4.92 2.57 2.25 1025 485.90 2.07 0.64 3.61 2.14 1.20
18 1072 438.67 1.74 0.64 4.65 2.09 0.74 910 254.16 1.62 0.43 3.80 1.64 0.43
19 1040 521.71 1.18 0.72 4.29 1.79 0.89 770 286.06 1.23 0.72 3.08 1.25 0.75
20 1242 583.85 1.57 0.72 4.51 1.96 1.04 955 1075.34 1.59 0.66 4.26 1.69 2.31
10/25/2017 Symbiosis International University & University of Windsor 115
116. CI with Linear and Exponential Probability for Multiple Rectangular Obstacles Case (MROC)
10/25/2017 Symbiosis International University & University of Windsor 116
CI with Linear Probability CI with Exponential Probability
Config.
Mean
(FE)
SD
(FE)
Mean
Time
(Seconds)
SD Time
(Seconds)
Mean
Distance
Travelled
Mean
Total Time
(Seconds)
SD Total
Time
(Seconds)
Mean
(FE)
SD
(FE)
Mean
Time
(Seconds)
SD Time
(Seconds)
Mean
Distance
Travelled
Mean
Total Time
(Seconds)
SD Total
Time
(Seconds)
1 1177 465.51 0.75 0.13 4.27 1.96 0.76 517 397.35 0.88 0.14 2.47 0.98 0.64
2 1220 560.11 0.85 0.13 4.39 2.28 0.98 630 361.54 1.01 0.23 3.06 1.25 0.56
3 1660 397.33 0.71 0.17 4.30 2.52 0.76 1115 531.20 0.88 0.17 2.99 2.04 0.80
4 1290 544.73 0.44 0.08 4.37 1.18 0.50 555 272.64 0.51 0.12 2.90 0.59 0.23
5 1265 532.74 0.63 0.14 4.64 1.55 0.75 575 358.62 0.71 0.20 2.88 0.96 0.52
6 1462 685.98 0.65 0.17 4.15 1.82 0.86 602 653.97 0.77 0.18 3.19 0.97 0.85
7 1465 487.58 0.73 0.23 4.04 2.16 0.80 1155 544.03 0.82 0.14 2.89 1.70 0.73
8 740 391.20 0.78 0.14 4.55 1.16 0.54 542 96.59 0.84 0.17 3.13 0.87 0.14
9 947 497.65 0.78 0.13 3.94 1.44 0.59 557.5 247.62 0.87 0.14 2.66 0.95 0.31
10 1377 502.45 0.82 0.16 4.15 1.82 0.71 632 196.63 0.95 0.18 2.50 1.09 0.20
11 1592 353.31 0.68 0.15 4.26 2.30 0.55 1045 367.17 0.76 2.29 2.73 1.15 2.46
12 1255 406.14 0.68 0.16 4.14 1.62 0.64 580 328.41 0.79 0.22 2.285 0.98 0.40
13 1165 380.96 0.64 0.18 4.25 1.57 0.60 545 384.64 0.70 0.17 2.60 0.80 0.62
14 1465 561.29 0.36 0.05 4.46 1.08 0.40 835 487.85 0.41 0.06 3.01 0.67 0.34
15 1280 556.98 0.36 0.06 4.03 0.96 0.39 652 357.15 0.43 0.06 2.72 0.54 0.24
16 1480 342.73 0.42 0.05 4.42 1.30 0.28 652 281.88 0.55 0.07 2.30 0.63 0.23
17 1517 612.94 0.44 0.06 4.15 1.38 0.52 662 436.64 0.53 0.06 2.69 0.66 0.35
18 1260 550.15 0.47 0.10 4.31 1.13 0.46 642 429.54 0.58 0.10 3.15 0.66 0.35
19 965 601.56 0.45 0.05 3.73 0.91 0.49 502 455.07 0.50 0.05 2.70 0.57 0.36
20 1447 502.71 0.42 0.07 4.30 1.29 0.42 730 351.28 0.51 0.07 2.98 0.71 0.28
117. CI with Linear and Exponential Probability for Cluttered Polygons Obstacles Case (CPOC)
10/25/2017 Symbiosis International University & University of Windsor 117
CI with Linear Probability CI with Exponential Probability
Config.
Mean
(FE)
SD
(FE)
Mean
Time
(Seconds)
SD Time
(Seconds)
Mean
Distance
Travelled
Mean
Total Time
(Seconds)
SD Total
Time
(Seconds)
Mean
(FE)
SD
(FE)
Mean
Time
(Seconds)
SD Time
(Seconds)
Mean
Distance
Travelled
Mean
Total Time
(Seconds)
SD Total
Time
(Seconds)
1 1085 381.58 1.18 1.89 4.70 1.73 2.01 827 350.53 1.17 0.36 3.76 1.45 0.52
2 1705 525.93 1.38 0.57 4.80 2.46 0.77 970 429.18 1.42 0.57 3.27 1.67 0.69
3 1335 529.86 1.45 1.47 4.09 2.05 1.70 770 477.73 1.34 0.29 2.91 1.48 0.59
4 1220 443.54 1.53 0.53 3.98 2.05 0.57 830 346.05 1.41 0.42 2.96 1.47 0.52
5 1212.5 369.66 1.68 0.45 4.23 2.24 0.66 810 658.66 1.47 0.38 3.40 1.57 1.29
6 1320 396.81 1.52 0.29 4.53 2.04 0.63 782 469.84 1.25 0.28 3.19 1.42 0.79
7 1245 460.44 1.35 0.36 3.84 1.92 0.92 797 457.86 1.25 0.35 3.05 1.46 0.79
8 1155 537.01 1.04 0.29 4.34 1.54 0.62 882 420.03 1.25 0.42 3.40 1.50 0.64
9 1105 373.17 1.13 0.47 3.80 1.61 0.68 782 187.49 1.21 0.82 2.64 1.23 0.84
10 1457 488.52 1.45 0.38 4.40 2.09 0.61 915 455.52 1.18 0.43 2.98 1.36 0.63
11 1012 329.10 1.73 0.70 3.83 2.09 0.80 905 201.53 1.86 0.46 3.36 1.91 0.51
12 1457 602.11 1.48 0.48 4.56 2.21 0.93 855 438.81 1.46 0.43 3.34 1.52 0.69
13 1067 609.75 1.61 0.47 4.28 2.07 1.09 840 591.33 1.48 0.66 3.64 1.80 1.26
14 1065 630.40 1.39 0.33 4.30 1.94 1.21 822 353.75 1.49 0.44 3.01 1.64 0.64
15 862 599.68 1.47 0.60 3.78 1.79 1.28 832 396.43 1.40 0.47 3.25 1.50 0.69
16 1832 669.64 1.65 0.49 4.90 3.29 1.36 1055 585.11 1.51 0.44 3.08 2.05 1.35
17 1587 465.60 1.90 0.63 4.11 2.76 0.79 1015 599.76 1.54 0.47 2.99 1.92 1.03
18 1237 650.33 1.30 0.42 4.45 2.10 1.45 820 340.38 1.51 0.43 2.87 1.64 0.51
19 1227 507.91 1.21 0.53 4.23 1.95 0.97 722 410.57 1.14 0.23 2.97 1.27 01.69
20 1430 684.84 1.25 0.61 4.18 2.21 1.54 937 538.51 1.30 0.50 2.84 1.90 0.88
118. The simulation shows that a
combination of different set of
obstacles and robots can be
solved by the program.
Cohort Intelligence can be
applied to enable the swarms to
move collectively towards the
source.
Collective intelligence is more
effective than individual
intelligence.
A perfect balance between linear
and high exponential probability
should be struck to avoid the
problem of local minima.
Exponential value has been taken
as 3 after observing trends of
both higher exponential and
linear probabilities.
Inference and Discussion
118
119. Cohort Intelligence can be effectively
used to solve complex real life tasks
such as search and rescue operations.
This project forms as a foundation for
the development of real life swarm
robots that would be capable of
achieving targets based on the principle
of Cohort Intelligence.
119
Project Conclusion and Future Direction
120. Project Conclusion and Future Direction
• In order to maintain a balance between collective learning and
individual intelligence, the exponential value 𝑧 was required to be
chosen based on preliminary trials. During, the preliminary trials it
was observed that for higher values 𝑧, more number of robots were
exhibiting individual intelligence rather than collective learning. As a
result, some of the robots got stuck behind the obstacles and could
not reach the target. In the near future, an approach to auto-tune
such parameter needs to be addressed. Also, CI with above two
approaches needs to be tested for complex U and V shaped obstacles.
Authors intend to apply reinforcement learning model for robots’
sapient systems
10/25/2017 Symbiosis International University & University of Windsor 120
121. Thank you
Anand J Kulkarni PhD, MASc, BEng, DME
Head & Associate Professor
Dept of Mechanical Eng
Symbiosis Institute of Technology
Symbiosis International University
Lavale, Pune 412 115, MH, India
Email: anand.kulkarni@sitpune.edu.in
kulk0003@ntu.edu.sg; kulk0003@uwindsor.ca; kulk0003@outlook.com
URL: sites.google.com/site/oatresearch/anand-jayant-kulkarni
Ph: 91 20 39116468; 91 7030129900
ResearcherID: www.researcherid.com/rid/O-3585-2016
ORCID ID: orcid.org/0000-0001-6242-9492
Google Scholar: scholar.google.ca/citations?user=IAvtDokAAAAJ&hl=en
10/25/2017 Symbiosis International University & University of Windsor 121