Exploratory Toolkit for Evolutionary and Swarm-Based Optimization

Abstract

Optimization of parameters or ’systems’ in general plays an ever-increasing role in mathematics, economics, engineering, and life sciences. As a result, a wide variety of both traditional analytical, mathematical and non-traditional algorithmic approaches have been introduced to solve challenging and practically relevant optimization problems. Evolutionary optimization methods~namely, genetic algorithms, genetic programming, and evolution strategies~represent a category of non-traditional optimization algorithms drawing inspirations from the process of natural evolution. Particle swarm optimization represents another set of more recently developed algorithmic optimizers inspired by social behaviours of organisms such as birds [8] and social insects. These new evolutionary approaches in optimization are now entering the stage, and are thus far very successful in solving real-world optimization problems [12]. Although these evolutionary approaches share many concepts, each one has its strengths and weaknesses. The best way to understand these techniques is through practical experience, in particular on smaller-scale problems or on commonly accepted benchmark functions. In [11], we describe how evolution strategies and particle swarm optimizers compare on benchmarks prepared for a much more complex optimization task regarding a kinematic model of a soccer kick. The Mathematica notebooks that we created throughout these evaluation experiments and for the final design of the muscle control algorithms for the soccer kick are now also available through a webMathematica interface. The new Evolutionary & Swarm Optimization web site is integrated with the collection of notebooks from the EVOLVICA package, which covers evolution-based optimizers from genetic algorithms and evolution strategies to evolutionary programming and genetic programming. The EVOLVICA database of notebooks, along with the newly added swarm algorithms, provide a large experimentation and inquiry platform for introducing evolutionary and swarm-based optimization techniques to those who either wish to further their knowledge in the evolutionary computation domain or require a streamlined platform to build prototypical strategies to solve their optimization tasks. Making these notebooks available through a web Mathematica site means that anyone with an internet browser available will have instant access to a wide range of optimization algorithms.