An improved PSO algorithm based on particle exploration for function optimization and the modeling of chaotic systems

A novel method to improve the global performance of particle swarm optimization (PSO) is proposed, which extends the exploring domain of the optimal position in the current generation and the optimal position thus achieved by every particle. In each generation, the best two positions are modified ac...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Soft computing (Berlin, Germany) Ročník 19; číslo 11; s. 3071 - 3081
Hlavní autoři: Chen, Debao, Chen, Jing, Jiang, Hao, Zou, Feng, Liu, Tundong
Médium: Journal Article
Jazyk:angličtina
Vydáno: Berlin/Heidelberg Springer Berlin Heidelberg 01.11.2015
Springer Nature B.V
Témata:
Algorithms
Artificial Intelligence
Computational Intelligence
Control
Engineering
Local optimization
Mathematical Logic and Foundations
Mechatronics
Methodologies and Application
Neighborhoods
Operators (mathematics)
Optimization
Particle swarm optimization
Performance enhancement
Robotics
Velocity
Linear decreasing inertia weight PSO (LDWPSO)
Chaotic system
Local convergence
Particle swarm optimization (PSO)
Ladder PSO
ISSN:1432-7643, 1433-7479
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:A novel method to improve the global performance of particle swarm optimization (PSO) is proposed, which extends the exploring domain of the optimal position in the current generation and the optimal position thus achieved by every particle. In each generation, the best two positions are modified according to their searching radii and directions. If the new solutions improve the old ones, the optimal positions in the updating equations of the conventional PSO algorithm will be replaced by the new solutions. Using this operator, the swarm diverges from local optimization easily. Moreover, the algorithm is easily implemented, and because the basic structure of PSO is not altered, the algorithm can be easily combined with different PSO methods to improve the performance. Some benchmark functions and chaotic systems are evaluated via simulations, showing that the proposed algorithm exceeds existing methods such as BPSO, LDWPSO, DNLPSO to some extent.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1432-7643
1433-7479
DOI:10.1007/s00500-014-1469-4