An Evo-Devo Framework for Computational Design Synthesis Based on a Regulation Process

Abstract

Evolutionary design synthesis has been used in solving many engineering design problems due to the reliability, robustness and domain-independence of evolutionary algorithms. However, little has been done to explore the role of design knowledge in evolutionary design synthesis. In this paper, inspired by recent advances in evolutionary developmental biology (Evo-Devo), we propose an Evo-Devo framework for computational design synthesis that utilizes design knowledge to guide an evolutionary search processes. The framework embodies a new regulation process to derive feasible phenotypes from the synthesis of genotypes and design knowledge. A multi-agent system (MAS) is implemented to perform the regulation needed to enforce the constraint satisfaction of phenotypes. The practical applicability of the Evo-Devo framework is validated through a meal design problem to be solved as a case study. The results indicate that the Evo-Devo framework and its regulation process outperform traditional processes in terms of effectiveness, efficiency and reliability in searching a constrained design space. The proposed framework and the MAS-based implementation of the regulation process can be applied to improve the performance of evolutionary design synthesis in different domains and tasks in general.