Researchers and practitioners in the field of evolutionary computation are the main audience for this book, but it can also be used easily for didactic activities for master’s and PhD students.
The book contains ten chapters, each one dealing with various aspects of the current state of the art in the field of evolutionary optimization algorithms applied to general constrained optimization problems. In addition, several interesting and challenging future research directions are highlighted.
Chapter 1 presents a survey of the adaptive penalty techniques applied in evolutionary computation. The penalty methods transform a constrained optimization problem within an unconstrained one, while an adaptive penalty method automatically sets the values of all the parameters involved using gathered information from the search process. The characteristics and applicability of different adaptive penalty methods used in the literature are presented, resulting in a better understanding of these techniques.
The effect of the fitness landscape on the performance of evolutionary algorithms is investigated in chapter 2, with a special emphasis on the importance of the infeasible regions of the search space. Chapter 3 describes a novel surrogate-assisted evolutionary programming algorithm for constrained expensive black-box optimization.
Chapter 4 investigates the ephemeral resource constraints introduced recently, which occur only temporarily during optimization in the case of some real-world applications. Their main features, in comparison to the classical soft and hard constraints, are: the solutions violating these constraints cannot be evaluated on the objective function, and usually these are functions of previously evaluated solutions. Several methods dealing with ephemeral resource constraints are described and evaluated.
An incremental approximation algorithm combined with a multi-membered evolutionary strategy is described in chapter 5. The idea of this novel approach is to approximate each constrained function with increasing accuracy. The computational results on a class of 13 benchmark problems prove the performance of the proposed approach.
In chapter 6, the authors propose a new scheme in order to use an approximation method for solving constrained optimization problems. While some computational results are reported, the proposed approach has not yet been applied to real-word optimization problems.
Chapter 7 deals with a conically constrained optimization problem and analyzes the behavior of multi-recombinative evolutionary strategies. Chapter 8 provides a novel strategy in order to locate feasible regions in constrained optimization problems by using a particle swarm optimization (PSO)-based technique.
In chapter 9, in the case of single-objective optimization problems, a new constrained handling technique consisting of four different constrained handling methods, each one having its own population, is described.
Finally, the last chapter presents an efficient hybrid algorithm obtained by combining in a complementary way a bi-objective evolutionary approach with a penalty function method.
This well-written book has several strengths. First, it deals with hot topics from the field of evolutionary constrained optimization. Second, it points out clearly the current state of the art in this research area. Finally, several challenging future research directions are presented. It will be useful for practitioners dealing with hard constrained optimization problems, as well as researchers and graduate students in the computer science and engineering fields.
