Edite Manuela da G. P. Fernandes

An augmented Lagrangian fish swarm based method for global optimization

This paper presents an augmented Lagrangian methodology with a stochastic population based algorithm for solving nonlinear constrained global optimization problems. The method approximately solves a sequence of simple bound global optimization subproblems using a fish swarm intelligent algorithm. A stochastic convergence analysis of the fish swarm iterative process is included. Numerical results with a benchmark set of problems are shown, including a comparison with other stochastic-type algorithms.

Hybridizing the electromagnetism-like algorithm with descent search for solving engineering design problems

In this paper, we present a new stochastic hybrid technique for constrained global optimization. It is a combination of the electromagnetism-like (EM) mechanism with a random local search, which is a derivative-free procedure with high ability of producing a descent direction. Since the original EM algorithm is specifically designed for solving bound constrained problems, the approach herein adopted for handling the inequality constraints of the problem relies on selective conditions that impose a sufficient reduction either in the constraints violation or in the objective function value, when comparing two points at a time. The hybrid EM method is tested on a set of benchmark engineering design problems and the numerical results demonstrate the effectiveness of the proposed approach. A comparison with results from other stochastic methods is also included.

Improved binary artificial fish swarm algorithm for the 0–1 multidimensional knapsack problems

The 0–1 multidimensional knapsack problem (MKP) arises in many fields of optimization and is NP-hard. Several exact as well as heuristic methods exist. Recently, an artificial fish swarm algorithm has been developed in continuous global optimization. The algorithm uses a population of points in space to represent the position of fish in the school. In this paper, a binary version of the artificial fish swarm algorithm is proposed for solving the 0–1 MKP. In the proposed method, a point is represented by a binary string of 0/1 bits. Each bit of a trial point is generated by copying the corresponding bit from the current point or from some other specified point, with equal probability. Occasionally, some randomly chosen bits of a selected point are changed from 0 to 1, or 1 to 0, with an user defined probability. The infeasible solutions are made feasible by a decoding algorithm. A simple heuristic add_item is implemented to each feasible point aiming to improve the quality of that solution. A periodic reinitialization of the population greatly improves the quality of the solutions obtained by the algorithm. The proposed method is tested on a set of benchmark instances and a comparison with other methods available in literature is shown. The comparison shows that the proposed method gives a competitive performance when solving this kind of problems.

Robot trajectory planning with semi-infinite programming

Abstract In this paper we describe how robot trajectory planning can be formulated as a semi-infinite programming (SIP) problem. The formulation as a SIP problem allowed us to treat the problem with one of the three main classes of methods for solving SIP, the discretization class. Two of the robotics trajectory planning problems formulated were coded in the SIPAMPL environment which is publicly available. A B-Spline library was also created to allow the codification of the robotics trajectory problem.

OPTIMIZATION OF NONLINEAR CONSTRAINED PARTICLE SWARM

We propose an algorithm based on the particle swarm paradigm (PSP) to address nonlinear constrained optimization problems. While some algorithms based on PSP have already been proposed in this context, the equality constraints have been posing some difficulties. The proposed algorithm is based on the relaxation of the dominance concept introduced in the multiobjective optimization. This concept is used to select the best particle position and the best ever particle swarm position. We propose also a stopping criterion for the algorithm and present numerical results with some problems collected from the literature. The new algorithm is implemented in a solver connected with AMPL, allowing easy coding and solving of problems.

Modified movement force vector in an electromagnetism-like mechanism for global optimization

This paper presents an algorithm for solving global optimization problems with bounded variables. The algorithm is a modification of the electromagnetism-like mechanism proposed by Birbil and Fang [An electromagnetism-like mechanism for global optimization, J. Global Optim. 25 (2003), pp. 263–282]. The differences are mainly on the local search procedure and on the force vector used to move each point in the population. Several widely-used benchmark problems were solved in a performance evaluation of the new algorithm when compared with the original one. A comparison with other stochastic methods is also included. The algorithm seems appropriate for large dimension problems.

Novel fish swarm heuristics for bound constrained global optimization problems

The heuristics herein presented are modified versions of the artificial fish swarm algorithm for global optimization. The new ideas aim to improve solution accuracy and reduce computational costs, in particular the number of function evaluations. The modifications also focus on special point movements, such as the random, search and the leap movements. A local search is applied to refine promising regions. An extension to bound constrained problems is also presented. To assess the performance of the two proposed heuristics, we use the performance profiles as proposed by Dolan and More in 2002. A comparison with three stochastic methods from the literature is included.

A hybrid genetic pattern search augmented Lagrangian method for constrained global optimization

Abstract Hybridization of genetic algorithms with local search approaches can enhance their performance in global optimization. Genetic algorithms, as most population based algorithms, require a considerable number of function evaluations. This may be an important drawback when the functions involved in the problem are computationally expensive as it occurs in most real world problems. Thus, in order to reduce the total number of function evaluations, local and global techniques may be combined. Moreover, the hybridization may provide a more effective trade-off between exploitation and exploration of the search space. In this study, we propose a new hybrid genetic algorithm based on a local pattern search that relies on an augmented Lagrangian function for constraint-handling. The local search strategy is used to improve the best approximation found by the genetic algorithm. Convergence to an e -global minimizer is proved. Numerical results and comparisons with other stochastic algorithms using a set of benchmark constrained problems are provided.

A reduction method for semi-infinite programming by means of a global stochastic approach

We describe a reduction algorithm for solving semi-infinite programming problems. The proposed algorithm uses the simulated annealing method equipped with a function stretching as a multi-local procedure, and a penalty technique for the finite optimization process. An exponential penalty merit function is reduced along each search direction to ensure convergence from any starting point. Our preliminary numerical results seem to show that the algorithm is very promising in practice. †This paper has been presented at the 8th International Conference on Parametric Optimization and Related Topics (Cairo, Egypt, November–December 2005).

SIPAMPL: Semi-infinite programming with AMPL

SIPAMPL is an environment for coding semi-infinite programming (SIP) problems. This environment includes a database containing a set of SIP problems that have been collected from the literature and a set of routines. It allows users to code their own SIP problems in AMPL, to use any problem already in the database, and to develop and test any SIP solver. The SIPAMPL routines support the interface between a potential SIP solver and test problems coded in AMPL. SIPAMPL also provides a tool that allows the selection of problems from the database with specified characteristics. As a concept demonstration, we show how MATLAB can use SIPAMPL to solve the problems in the database. The Linux and Microsoft Windows versions together with the database of coded problems are freely available via the web.