Object oriented toolkit for multiobjective genetic optimisation

Abstract

Evolutionary computations are emerging as powerful tools for search and optimisation, and increasingly being used in many scientific and engineering applications. Side-by-side, object oriented computing has revolutionised, during the current decade, the style of programming and software system design and development which is now configured around the 'class' concept. We present a general purpose object oriented toolkit which serves as a generic utility for wide ranging applications involving optimisation of both single and multiple objectives. The toolkit supports the state of the art of genetic optimisation techniques; the design is modular, flexible and extensible, in line with the object oriented programming paradigm. The toolkit is currently being implemented in C++ for obvious reasons of wider support and portability across platforms. Templates and derived classes are used for elegance and reuse of the code and the library. The interfaces try to hide as many of the implementation details as possible so that the programming and modification at higher level become simple. Nonetheless, defining interfaces is an iterative process, so with the design and implementation of the toolkit, with each major addition and upgrade, they are constantly evolving.