Comparing the performance of evolutionary algorithms for permutation constraint satisfaction

Abstract:

This paper presents a systematic comparison of canonical versions of two evolutionary algorithms, namely Particle Swarm Optimization (PSO) and Genetic Algorithm (GA), for permutation constraint satisfaction (permut-CSP). Permut-CSP is first characterized and a test case is designed. Agents are then presented, tuned and compared. They are also compared with two classic methods (A*and hill climbing). Results show that PSO statistically outperforms all other agents, suggesting that canonical implementations of this technique return the best trade-off between performance and development cost for our test case. © 2011 Authors.

Año de publicación:

2011

Keywords:

• csp
• pso
• constraint satisfaction
• Genetic Algorithm

Fuente:

scopus