[Paper Review] Hybrid Model for Solving Multi-Objective Problems Using Evolutionary Algorithm and Tabu Search

Rjab Hajlaoui, Mariem Gzara+1 more|arXiv (Cornell University)|Feb 15, 2011

Advanced Multi-Objective Optimization Algorithms6 references5 citations

TL;DR

This paper proposes a hybrid multi-objective optimization model combining evolutionary algorithms (EA) for global exploration and tabu search (TS) for local intensification. The approach improves convergence and diversity on ZDT benchmark problems by leveraging EA's population-based search and TS's memory-driven local search, achieving superior Pareto front approximation compared to standalone methods.

ABSTRACT

This paper presents a new multi-objective hybrid model that makes cooperation between the strength of research of neighborhood methods presented by the tabu search (TS) and the important exploration capacity of evolutionary algorithm. This model was implemented and tested in benchmark functions (ZDT1, ZDT2, and ZDT3), using a network of computers.

Motivation & Objective

• To address the challenge of balancing convergence and diversity in multi-objective optimization.
• To integrate the global search capability of evolutionary algorithms with the local search efficiency of tabu search.
• To develop a cooperative hybrid framework that enhances solution quality and convergence speed.
• To evaluate the model on standard multi-objective test functions (ZDT1, ZDT2, ZDT3).
• To demonstrate the effectiveness of the hybrid approach in achieving a well-distributed and converged Pareto front.

Proposed method

• The hybrid model uses an evolutionary algorithm to generate an initial population of solutions for multi-objective problems.
• Tabu search is applied iteratively to improve individual solutions by exploring neighboring solutions while avoiding cycles via memory structures.
• The algorithm alternates between evolutionary operations (selection, crossover, mutation) and tabu search refinement to enhance solution quality.
• A shared external archive maintains non-dominated solutions, guiding the search toward the true Pareto front.
• The cooperation mechanism allows the EA to explore diverse regions while TS refines promising solutions locally.
• Computational experiments were conducted on a computer network to evaluate performance on ZDT benchmark functions.

Experimental results

Research questions

• RQ1 How can the global exploration of evolutionary algorithms be effectively combined with the local intensification of tabu search in multi-objective optimization?
• RQ2 To what extent does the hybrid model improve convergence and diversity of the Pareto front compared to standalone EA or TS?
• RQ3 Can the memory-based local search of tabu search enhance the convergence speed of evolutionary algorithms in multi-objective settings?
• RQ4 How does the hybrid model perform on standard multi-objective test functions such as ZDT1, ZDT2, and ZDT3?
• RQ5 What is the impact of the cooperation mechanism between EA and TS on the quality and distribution of non-dominated solutions?

Key findings

• The hybrid model achieved better convergence and diversity metrics on ZDT1, ZDT2, and ZDT3 compared to individual EA and TS approaches.
• The integration of tabu search significantly improved the convergence rate of the evolutionary algorithm on all benchmark functions.
• The external archive maintained a well-distributed set of non-dominated solutions, indicating effective diversity preservation.
• The model demonstrated robust performance across different problem characteristics, including convex and non-convex Pareto fronts.
• The computational experiments confirmed that the cooperation between EA and TS enhances both convergence and spread of the Pareto front.
• The hybrid approach outperformed state-of-the-art single-method approaches on the tested multi-objective problems.