Abstract:Engineering design optimization poses a significant challenge, usually requiring human expertise to discover superior solutions. Although various search techniques have been employed to generate diverse designs, their effectiveness is often limited by problem-specific parameter tuning, making them less generalizable and scalable. This article introduces a framework inspired by evolutionary and developmental (evo-devo) concepts, aiming to automate the evolution of structural engineering designs. In biological systems, evo-devo governs the growth of single-cell organisms into multicellular organisms through the use of gene regulatory networks (GRNs). GRNs are inherently complex and highly nonlinear, and this article explores the use of neural networks and genetic programming as artificial representations of GRNs to emulate such behaviors. To evolve a wide range of Pareto fronts for artificial GRNs, this article introduces a new technique, a real value-encoded neuroevolutionary method termed real-encoded NEAT (RNEAT). The performance of RNEAT is compared with that of two well-known evolutionary search techniques across different 2-D and 3-D problems. The experimental results demonstrate two key findings. First, the proposed framework effectively generates a population of GRNs that can produce diverse structures for both 2-D and 3-D problems. Second, the proposed RNEAT algorithm outperforms its competitors on more than 50% of the problems examined. These results validate the proof of concept underlying the proposed evo-devo-based engineering design evolution.

Evolving Novel Gene Regulatory Networks for Structural Engineering Designs

Neural Graph Evolution: Towards Efficient Automatic Robot Design

Neuroevolution of Recurrent Architectures on Control Tasks

Evolving Differentiable Gene Regulatory Networks

A novel generative reverse net assisted evolution algorithm for expensive-computational optimizations

Evo-Devo Algorithms: Gene-Regulation for Digital Architecture

Born to Learn: the Inspiration, Progress, and Future of Evolved Plastic Artificial Neural Networks

Dynamic Links and Evolutionary History in Simulated Gene Regulatory Networks

Functional generative design of mechanisms with recurrent neural networks and novelty search

Searching Search Spaces: Meta-evolving a Geometric Encoding for Neural Networks

Efficient and Diverse Generative Robot Designs using Evolution and Intrinsic Motivation

Efficient automatic design of robots

NeuroEvolution algorithms applied in the designing process of biohybrid actuators

A Cellular Mechanism for Multi-Robot Construction Via Evolutionary Multi-Objective Optimization of a Gene Regulatory Network

EvoRobogami: Co-designing with Humans in Evolutionary Robotics Experiments

NeuroEvo: A Cloud-based Platform for Automated Design and Training of Neural Networks using Evolutionary and Particle Swarm Algorithms

The Networked Evolutionary Algorithm: A Network Science Perspective

Neuroevolution in Deep Neural Networks: Current Trends and Future Challenges

A multi-objective differential evolutionary approach toward more stable gene regulatory networks

Protein pathways as a catalyst to directed evolution of the topology of artificial neural networks

Evolving Modular Genetic Regulatory Networks with a Recursive, Top-Down Approach