Evolution and Optimum Seeking: The Sixth Generation
Evolution and Optimum Seeking: The Sixth Generation
Approximating the Nondominated Front Using the Pareto Archived Evolution Strategy
Evolutionary Computation
Hi-index | 0.00 |
The Pareto Archived Evolution Strategy (PAES), one of the most successful evolutionary optimizers, has long been proven as an easy-to-implement algorithm due to its simple (1+1) search to solve for multiobjective problems. However, a recent comparative study with Multiobjective Simulated Annealing (MOSA), another potential heuristic search technique, showed that when the problems are constrained or becoming more complex, e.g. with a large number of control variables, PAES seemed not to explore the trade-off surface satisfactorily. By examining the nature of MOSA, this paper attempts to improve the performance of PAES by adding the sensitivity adjustment, one of the key characteristics of MOSA implementation. Based on 4 standard test problems with either a large number of control variables or with three or more objectives, comparative results indicate that the performance of the PAES algorithm with the addition of sensitivity adjustment has been improved significantly. In one test problem, the performance of PAES even outperforms that of MOSA. On-going research is on progress to extend the test covering a wide range of different complex optimisation problems.