Evolution is basically a biological optimization algorithm

  • The problem
  • In daily practice, natural scientists and engineers often have to deal with optimization problems for which no analytical descriptions are known and / or for which the number of variables to be optimized at the same time is very large or the contextual relationships are highly non-linear. Examples of this can be found in fields as diverse as jet engine design, the automotive industry, and particle physics. In the Bionics - an artificial word from biology and technology - the solutions, structures and principles found in nature are examined to see whether they represent optimal solutions and to what extent they can be used for new concepts in the optimization of technical problems.

  • Examples in nature

  • That optimal solutions exist in nature is suggested e.g. by so-called allometric laws for vertebrates. Further examples are the optimal branching structure of the blood vessel system in humans, the garland flight of birds and the optimal structure of the complex eye of insects. In many of the examples given, it is possible to prove analytically by means of differential calculus methods that these are actually optimal solutions in nature. Among the newer methods that come into question for the above-mentioned optimization problems, those that emulate biological evolution have a special position because they can also be used to deal with highly complex industrial problems.

  • The principle of evolutionary strategy

  • The evolution strategy, thirty years ago as an optimization method by Prof. Dr. Ingo Rechenberg introduced a universal procedure for solving and optimizing technical problems. There is a very strong distinction between the contents of the terms evolution strategy and evolution technology. In contrast to the genetic algorithms, the first term subsumes the optimization processes or strategy variants that are based on the events of natural evolution. In recent times it has become established to collectively refer to evolution strategies, genetic algorithms and genetic programming as evolutionary algorithms.

  • Evolution technology

  • On the other hand, evolution technology means the methods and heuristics of problem processing in the sense of a problem-relevant formulation of the quality function necessary for every optimization process. In addition, in addition to the actual optimization problem, additional conditions and compliance with manufacturing tolerances must usually be taken into account, especially in the case of industrial issues.

  • implementation

  • So how can the principles on which optimization is based in the context of biological evolution be used for the search for optimal solutions in the design of technical systems? Newly developed algorithms are used which, among other things, imitate the mechanisms of evolution such as mutation, selection and recombination. In addition, quality functions adapted to the problem must be derived, which allow the consideration of secondary conditions and tolerances. The consideration of secondary conditions in the form of specially constructed penalty function components of the quality function for the optimization by means of evolution strategies is an essential milestone. With this approach it is often only possible to generate a valid starting solution in the high-dimensional variable space.

  • Modeling problems

  • The modeling effort that has to be made in order to make a problem accessible to optimization at all can in some cases be quite considerable. In the context of the evolution strategy, one speaks of specifying or more precisely "constructing" a quality function through suitable modeling of the problem so that the strong causality is fulfilled. Such quality functions are characterized by small changes in the case of small changes in the variables. The problem to be optimized is usually characterized by the (physical) laws describing the problem and the associated boundary and secondary conditions. In general, the problem to be considered cannot be accessed by the optimization without further processing. Optimization by means of a selected method is only possible after suitable modeling, which is almost always accompanied by a simplification and reformulation of the original optimization task. The evolution strategy, as an adaptive stochastic search method, is one of many methods to process or solve optimization problems. Why does it now occupy a special position among the variety of optimization methods? It can be used both as a mathematical optimization method and in the laboratory for the experimental setting of optimal variable settings.

  • outlook

  • The outstanding position of this optimization method is due to the fact that the principles used, such as B. Mutation, recombination and selection have been used for millions of years in the context of biological evolution for the optimal adaptation of structures, processes and interactions in animate and inanimate nature. Since Darwin (1809-1882) first described the mechanisms underlying natural evolution, the theory of evolution has matured into one of the best explanatory models for understanding the variety of phenomena in nature. In addition, it is of decisive importance that the evaluation and assessment of the realizations derived from a basic structure can be carried out in parallel. In industrial applications in particular, the time savings that this saves in optimization cannot be valued highly enough, especially in view of the ever shorter product life cycles.

    Last change: 01/08/01 by Norbert Kalus