Repeated weighted boosting search for discrete or mixed search space and multiple-objective optimisation

Repeated weighted boosting search (RWBS) optimisation is a guided stochastic search algorithm that is capable of handling the difficult optimisation problems with non-smooth and/or multi-modal cost functions. Compared with other alternatives for global optimisation solvers, such as the genetic algorithms and adaptive simulated annealing, RWBS is easier to implement, has fewer algorithmic parameters to tune and has been shown to provide similar levels of performance on many benchmark problems. In its original form, however, RWBS is only applicable to unconstrained, single-objective problems with continuous search spaces. This contribution begins with an analysis of the performance of the original RWBS algorithm and then proceeds to develop a number of novel extensions to the algorithm which facilitate its application to a more general class of optimisation problems, including those with discrete and mixed search spaces as well as multiple objective functions. The performance of the extended or generalised RWBS algorithms are compared with other standard techniques on a range of benchmark problems, and the results obtained demonstrate that the proposed generalised RWBS algorithms offer excellent performance whilst retaining the benefits of the original RWBS algorithm. © 2012 Elsevier B.V. All rights reserved.