#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Alaya 30 10
1. Distributed Genetic Algorithm NSGA II
for solving the DARP
Alaya raddaoui1
and Kamel zidi2
1 alayaraddaoui@gmail.com
2 kamel_zidi@yahoo.fr
2. SOIE PresentationSOIE Presentation
Stratégies d’Optimisation et Informatique intelligentE
-Research thematic :
knowledge Engineering and reasoning:
Reasoning and Optimization under Constraints
Multi-Agent Systems
Systems, information and services engineering:
Systems and Information Engineering
Services Engineering
-Adress:
Laboratoire de recherche Stratégies d’Optimisation et Informatique intlligentE SOIE
ISG Tunis, 41, Rue de la Liberté, Cité Bouchoucha 2000 Le bardo, Tunis-TUNISIE 2
3. Outline
2. Problematic
3. State of the art
4. Proposed approach
5. Obtained results
6. Conclusion and perspectives
1. Introduction
3
4. 4
Transport problems can have effects on the environment at
different levels:
Global;
regional ;
local.
[ONU 2001]
Introduction
5. 5
The improved transport system seems to be a necessity
because its complexity is a reality. This system is also
affected by the following phenomena:
Social;
Economic;
Structural.
Introduction
6. S4
S7
S2
S3
S1
S8
S5
S6
: Passengers
S : Station
: Vehicle
Parametres (departure T,
arrival T...)
: Destination
Optimise the tours of vehicles to answer
the passengers requests 6
Problematic
Dial a Ride Problem: DARP
7. 7
DARP resolution
(Psaraftis, 1980) (Cordeau&laporte,2003) (Stefan, 2005) (Mauri et al,2006) (Claudio et al,2009) (Zidi et al,10)
Exact algorithm:
dynamic
programming
Taboo search
algorithm
Branch and
Bound
method
Simulated
annealing
algorithm
Genetic
algorithm
Multiobjective
simulated
annealing
algorithm
State of the art(DARP)
8. 8
Presentation
Origin : Darwin's theory of evolution
Coding chromosomal structures
natural selection
Evolution operators
Selection
Crossing
Mutation [Goldberg 89]
State of the art(GA)
10. 10
Genetic Algorithm NSGA2
Presentation
NSGAII (Elitist Non-dominated Sorting Genetic Algorithm)
Proposed by Deb and his team[2000]
Based on three characteristics:
The principle of elitism
The non-dominated solutions
Variety of explicit solutions
[Deb and 2000]
State of the art(GA)
12. Interface agent:
- Generate randomly the initial population.
- Create species agents for each sub-population.
- Create new agents species if they exist.
- Detect the best partial solution.
Specie agent:
- Execute his own distributed genetic algorithm.
Proposed approach
Our multi-agents architecture
12
13. Distribution of NSGA2
13
Interface Agent
initial population Evaluation
Rank1 Rank2 Rank3 … Rank n
Non dominated sorting
…
Sélection agent Crossing agent Mutation agent
Sélection agent Crossing agent Mutation agent
Sélection agent Crossing agent Mutation agent
Species1 agent
Species3 agent
Species2 agent
Proposed approach
14. 14
The distributed genetic algorithm NSGA2
Creation of initial population (cities, deposits, connection ...)
Sort by rank
Do
Creating an agent for each species rank
Launch the local genetic algorithm to each agent species
Exchange of individuals crossing
Exchange of new individuals
Wihle (Number of generations reached)
Proposed approach
15. Local genetic algorithm for species agents
1- Crossover of the selected sub-population.
2- Update the obtained sub-population (Child).
3- Mutation of the sub-population child crossed.
4- Update the mutated sub-population child.
Proposed approach
16. 16
duration of the road according to the number of requests
Instance1 (24
requests)
Instance2 (36
requests)
Instance3
(48
requests)
Instance4 (72
requests)
Instance5
(120
requests)
AGD(NSGAII)
1249,15
6 vehiculs
2150,46
8 vehiculs
4003.95
8 vehiculs
RSMO (Zidi
et all 10) 1414,38
3 vehiculs
1407,6
8 vehiculs
1808,99
11 vehiculs
2270,86
4O20 ,75
13 vehiculs
1436,23
3 vehiculs
1404,4
4 vehiculs
Obtained Results
17. 17
Execution time based on the number of requests
Instance1
(24
requests)
Instance2
(36
requests)
Instance3
(48 requests)
Instance4
(72requests)
Instance4
(120requests)
DGA(NSGAII) 0 ,71 2,88 3 ,08 3,86 6,77
RSMO (Zidi
et all 10)
0,57 2,32 4,70 4,90 9,61
Obtained Results
18. 18
Duration of the road according to the number of
requests
Execution time based on the number of requests
Efficiency and improving 4/5 times (duration of the
road) and 3/5 times (run of time)
Obtained Results
19. 19
Modeling of DARP with two objectives
Resolution of DARP for Distributed genetic algorithm NSGA-II
Use of multi-agent system approach to distribution of the algorithm
NSGA II
Conclusion
Conclusion and perspectives
20. 20
Application of the approach on real data.
Hybridization of DGA NSGA II with other accurate methods and
algorithms.
Perspectives
Conclusion and perspectives