Comparison of Two Multi-Agent Algorithms: ACO and PSO for the Optimization of a Brushless DC Wheel Motor
Particle swarm optimization (PSO) and ant-colony optimization (ACO) are novel multi-agent algorithms able to solve complex problem. By consequence, it would seem wise to compare their performances for solving such problems. For this purpose, both algorith
- PDF / 220,668 Bytes
- 8 Pages / 439 x 666 pts Page_size
- 14 Downloads / 280 Views
Abstract Particle swarm optimization (PSO) and ant-colony optimization (ACO) are novel multi-agent algorithms able to solve complex problem. By consequence, it would seem wise to compare their performances for solving such problems. For this purpose, both algorithms are compared together and with Matlab’s GA in term of accuracy of the solution and computation time. In this paper the optimization is applied on the design of a brushless DC wheel motor that is known as a nonlinear multimodal benchmark.
1 Introduction Often, to shown the practical interest of any optimization metaheuristic it is necessary to test them through more and more difficult experiments like solving hard optimization problems. For improving its applicability, it is important to answer how and why the methods work. This paper tackles the comparison of the two natureinspired metaheuristics particle swarm optimization (PSO) and ant-colony optimization (ACO) in order to adapt them for handling hard optimization problems. Indeed, PSO and ACO are novel techniques created for solving hard optimization problems. As they are multi-agent methods, they can be used to handle multi-level optimization problems. Unfortunately, the tuning of the control parameters for these algorithms is not acquired. So, to do a successful selection of their control parameters, a brushless DC wheel motor benchmark is used. In this study, the GA algorithm implemented in the Matlab Genetic Algorithm and Direct Search Toolbox is used to evaluate both multi-agents algorithms. The comparison among the both algorithms and the brushless DC wheel motor benchmark are presented in Sects. 1 and 2, respectively. In Sect. 3, the mixed-integer F. Moussouni, S. Brisset, and P. Brochet L2EP – Ecole Centrale de Lille, Cit´e Scientifique, BP 48, 59651 Villeneuve d’Ascq Cedex, France [email protected] F. Moussouni et al.: Comparison of Two Multi-Agent Algorithms: ACO and PSO for the Optimization of a Brushless DC Wheel Motor, Studies in Computational Intelligence (SCI) 119, 3–10 (2008) c Springer-Verlag Berlin Heidelberg 2008 www.springerlink.com
4
F. Moussouni et al.
NSGA-II is applied to tune ACO and PSO parameters in order to reduce the Euclidean distance between the known optimal point and the solution found by these algorithms, and the number of evaluations. Then, both methods (ACO and PSO) are evaluated in terms of convergence speed, and quality of the results in Sect. 4. Thereafter Sect. 5 tackles how the both algorithms are behaving to obtain effective solutions to a multi-level optimization problem. This is one open question among many with a certain interest of being solved in the near future. The final section, gives some concluding remarks.
2 Comparing PSO and ACO Methods PSO and ACO are considered as a global optimization method. In particular, these algorithms manage very well combinatorial and mixed problems. Unlike gradient search methods, they are less susceptible to be trapped in local optima. When GA mimics the natural biological evolution, ACO and PSO draw ins
Data Loading...
