Multi-objective Optimization: Bio-mimetic Adaptations of Genetic Algorithm

Abstract

Optimization using more than one objective function (multi-objective optimization) has become quite common in engineering, since it provides a decision-maker with an idea of the trade-offs between non-commensurate objectives. An example is the maximization of the profit and minimization of pollution. Pareto optimal sets of equally good (non-dominated) solutions are obtained from which the “preferred solution/operating point” has to be selected, often using intuition. Evolutionary techniques, e.g. genetic algorithm and simulated annealing, have become quite popular since they are extremely robust. However, they take considerable computational effort and any adaptation to speed up the rate of convergence is welcome. Two ideas from biology, namely jumping genes (transposons) and the altruism of honey bees, have been bio-mimicked to give faster adaptations of multi-objective genetic algorithms as well as simulated annealing. Two compute-intense applications in chemical engineering using these algorithms are discussed.

Keywords:

• Altruism
• Genetic algorithm
• Jumping genes
• Multi-objective optimization
• Pareto set

Acknowledgement

This material was presented as the Deepak Group's Padma Bhushan Professor L. K. Doraiswamy Chemcon Distinguished Speaker Award, Chemcon-2010, Annamalai University, Annamalainagar, 27–29 December 2010.