[Paper Review] Tracking Extrema in Dynamic Environment using Multi-Swarm Cellular PSO with Local Search

Somayeh Nabizadeh, Alireza Rezvanian+1 more|arXiv (Cornell University)|Jul 31, 2013

Metaheuristic Optimization Algorithms Research23 references10 citations

TL;DR

This paper proposes a multi-swarm cellular particle swarm optimization (PSO) algorithm with local search to track optima in dynamic environments. By partitioning the search space into cells, forming sub-swarms through clustering, and applying local search per cell, the method improves convergence and adaptability, outperforming alternatives on both static benchmarks and dynamic test problems.

ABSTRACT

Many real-world phenomena can be modelled as dynamic optimization problems. In such cases, the environment problem changes dynamically and therefore, conventional methods are not capable of dealing with such problems. In this paper, a novel multi-swarm cellular particle swarm optimization algorithm is proposed by clustering and local search. In the proposed algorithm, the search space is partitioned into cells, while the particles identify changes in the search space and form clusters to create sub-swarms. Then a local search is applied to improve the solutions in the each cell. Simulation results for static standard benchmarks and dynamic environments show superiority of the proposed method over other alternative approaches.

Motivation & Objective

• Address the challenge of tracking optimal solutions in dynamic optimization problems where the environment changes over time.
• Overcome limitations of conventional optimization methods that fail to adapt to shifting optima.
• Improve search efficiency and solution quality in dynamic environments through localized search and swarm decomposition.
• Enhance robustness and convergence speed by leveraging spatial partitioning and sub-swarm formation.
• Demonstrate superiority over existing PSO variants in dynamic benchmark environments through simulation.

Proposed method

• Partition the search space into a grid of cells to enable localized optimization and spatial awareness.
• Use particle clustering within each cell to form sub-swarms, enabling localized information sharing and coordinated search.
• Apply a local search strategy within each cell to refine solutions and improve convergence speed.
• Dynamically detect environmental changes through particle monitoring and trigger sub-swarm reformation when necessary.
• Integrate cellular topology to maintain diversity and prevent premature convergence.
• Combine global exploration via multi-swarm dynamics with local exploitation via cell-specific local search.

Experimental results

Research questions

• RQ1 Can a cellular PSO framework with sub-swarm formation improve tracking of optima in dynamic environments?
• RQ2 How does the integration of local search within cells affect convergence speed and solution quality in dynamic optimization?
• RQ3 To what extent does the proposed method outperform standard PSO and other multi-swarm variants in dynamic benchmark problems?
• RQ4 How effective is the clustering mechanism in adapting to environmental changes and maintaining solution diversity?
• RQ5 What is the impact of spatial partitioning on the robustness and scalability of the algorithm?

Key findings

• The proposed algorithm achieves superior performance on standard static benchmarks compared to conventional PSO and other multi-swarm variants.
• In dynamic environments, the method demonstrates enhanced ability to track moving optima due to adaptive sub-swarm formation and local search.
• The integration of local search significantly improves convergence speed and solution accuracy in each cell.
• The cellular structure enhances diversity and reduces the risk of premature convergence during optimization.
• Simulation results confirm the method's robustness across multiple dynamic test functions, outperforming existing approaches.
• The algorithm effectively detects environmental changes and responds by reorganizing sub-swarms to maintain tracking capability.