A Multi-Agent Systems Based Conceptual Ship Design Decision

A Multi-Agent Systems Based Conceptual Ship Design
Decision Support System
The Ship Stability Research Centre
Department of Naval Architecture and Marine Engineering
Universities of Glasgow and Strathclyde
Bekir S. Türkmen
Motivations
•
Design Exploration and Support
•
Distributed Architecture
•
Encapsulation of Design Experience
What is an agent?
An Agent : one that acts or has the power or authority to
act or represent another.
An Intelligent Agent is the agent does the things
rationally in a given situation (Russell 1995)
Intelligent Agents
•
•
•
•
•
•
Autonomy
Collaborative Behaviour
Adaptivity
Mobility
Proactivity
Reactivity
Multi-Agent Systems
MAS- Three Important Questions
•
Communication
•
Control
•
Co-ordination, Collaboration, Negotiation
Communication
Semantics and Syntax
FIPA-ACL
(INFORM
• KQML, FIPA-ACL
• KIF, FIPA-SL
:sender ( agent-identifier :name Sender@BEKIRN:1099/JADE
:addresses ()
:receiver (set ( agent-identifier :name
Receiver@BEKIRN:1099/JADE) )
:content "Hello SSRC"
KQML/KIF
)
(evaluate
FIPA-SL
:sender A :receiver B
(query-ref
:language KIF :ontology motors
:sender (agent-idenfier :name B)
:reply-with q1 :content (val (torque m1)))
(reply
:sender B :receiver A
:language KIF :ontology motors
:in-reply-to q1 :content (= (torque m1) (scalar 12 kgf)))
:receiver (set (agent-identifier :name
A))
:content
((iota ?x (p ?x)))
:language FIPA-SL
:reply-with query1)
Control
•
Centralized
•
Federated
•
Autonomous
Co-ordination
Auctions
Contract-Net (Task Sharing)
Planning
Game Theory
Argumentation
Catalogue of Conflicts
Proposed IA Architecture
ENVIRONMENT
Acquaintance Module
Communication Layer
User Interface
Coordination Layer
•
List of Agents
•
Agents’ work definition
Knowledge Base for Conflicts
•
Rule-based
•
Case-based
Acquaintance Module
Conflict Resolution Module
Optimisation Module
Optimisation Module
•
Local-Search Algorithms
•
Global-Search Algorithms
Learning Module
Task Layer
Task Layer
Intelligent Agent Architecture
•
Knowledge Base
•
Wrapped Simulation Tools
Proposed MAS Architecture
Worker Agents
Decision Theoretic Agents
CFD Agent
Multi-Attribute Decision
Maker Agent
Static Stability Agent
Dynamic Stability Agent
Multi-Objective
Optimisation Agent
Evacuation Agent
Resistance Agent
Hull Generation Agent
FEA Agent
………………………..
Multi-Agent System Architecture
Geometry
Transfer
User Interface Agents
3D Real-Time Simulation /
Virtual Reality Agent
Decision-Theoretic Agents
Decision Theoretic Agents
Multi-Attribute Decision
Maker Agent
Ranking and Selection Methods
TOPSIS (Technique for Order Preference
by Similarity to Ideal Solution)
……
Multi-Objective
Optimisation Agent
Multi-Objective Optimisation Algorithms
VEGA (Vector Evaluated GA)
NSGA (Non-Dominated Sorting GA)
NSGA2 (A Fast and Elitist NSGA)
SPEA/SPEA2 ( Strength Pareto Genetic
Algorithm)
Multi-Objective Optimisation
•
Decision-Making Before Search
•
Decision-Making After Search
•
Decision-Making during Search
Comparison of MOGA Methods
Figure 1
4
4
3.5
3.5
3
3
2.5
2.5
2
2
1.5
1.5
1
1
0.5
0.5
0
0
0
1
2
3
4
f1
0
5
1
2
3
Figure 3
4.5
f2
Figure 2
4.5
f2
f2
4.5
Objective Functions :
4
3.5
f1(x) = x2
3
; f2(x) = (x-2)2
Figure 1. VEGA Results
2.5
2
Figure 2. NSGA Results
1.5
1
Figure 3. NSGA II Results
0.5
0
0
1
2
3
f1
4
4
f1
5
Integrated Decision-Making and Search
In order to reduce the calculation cost and scalability we guide the
search by introducing designer preferences into search.
• Applied as A Priori and Progressive,
• Final Selection from Reduced Pareto-Set
f2
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
0
1
2
3
4
f1
5
Proposed Approach for Introducing Bias
•NSGA II + TOPSIS Algorithm
• Reference Point Method Approach
f2
4.5
4
NADIR
POINT
3.5
3
2.5
2
1.5
1
0.5
0
0
IDEAL
POINT 1
2
3
4
f1
5
Proposed Approach for Introducing Bias
Continued
Two modifications to introduce bias,
• Modification of Elitist Strategy
• Modification of Crowding Distance Assignment
Preference is given as, one unit of a is worth at most x units of b
1.2
NoBias



f1 (x)  x1

2 
  x  
f 2 (x)  g (x) 1   1    xi  0,1
  g (x)   

n

x

i

i 2
g (x)  1  9.

n 1

w1= 0.5
1
F2
0.8
0.6
0.4
0.2
0
0
0.2
0.4
0.6
F1
0.8
1
1.2
Internal Hull Subdivision Optimisation
Objectives
• Survivability –Max.
• Cargo Capacity (In Car Lanes) Max.
• Limiting KG – Max.
Constraints
Two Adjacent Bulkhead Distance greater than SOLAS’90 Longitudinal
Damage Extent,
SOLAS’ 90 Regulations,Limiting KG Reduction for operational Life cycle
Internal Hull Subdivision Optimisation
Internal Hull Subdivision Optimisation
Continued
14.14
14.12
Limiting KG
14.1
NoBias
14.08
Kg Important
14.06
Hs Important
14.04
14.02
14
0
2
4
6
8
10
12
Cargo
Hold Capacity
Cargo
Capacity
(Car Lanes)
14
Internal Hull Subdivision Optimisation
Results
Distributed Optimisation Test Problem in A
Multi-Agent Systems
Distributed Optimisation in A MAS
6
Early Results
5
4
3
2
1
0
0
1
2
3
4
5
6
Conclusions and Future Development
Advantages of proposed approach
• Distributed Computation (Less computation time)
• Distribution of Expertise (Intelligent Agent Architecture)
• Integrated Multi-Criteria Decision-Making and Decision
Support Environment.
Future Research
Integration with CAD Environment
Case Study for Intelligent Agents in Multi-Agent Systems
Questions