Learning to Design From Humans: Imitating Human Designers Through Deep Learning

Abstract

Humans as designers have quite versatile problem-solving strategies. Computer agents on the other hand can access large scale computational resources to solve certain design problems. Hence, if agents can learn from human behavior, a synergetic human-agent problem solving team can be created. This paper presents an approach to extract human design strategies and implicit rules, purely from historical human data, and use that for design generation. A two-step framework that learns to imitate human design strategies from observation is proposed and implemented. This framework makes use of deep learning constructs to learn to generate designs without any explicit information about objective and performance metrics. The framework is designed to interact with the problem through a visual interface as humans did when solving the problem. It is trained to imitate a set of human designers by observing their design state sequences without inducing problem-specific modelling bias or extra information about the problem. Furthermore, an endto-end agent is developed that uses this deep learning framework as its core in conjunction with image processing to map pixel-todesign moves as a mechanism to generate designs. Finally, the designs generated by a computational team of these agents are then compared to actual human data for teams solving a truss design problem. Results demonstrates that these agents are able to create feasible and efficient truss designs without guidance, showing that this methodology allows agents to learn effective design strategies.

TopologyGAN: Topology Optimization Using Generative Adversarial Networks Based on Physical Fields Over the Initial Domain

“…In recent years, new data-driven methods for topology optimization have been proposed to accelerate the process. Deep learning methods have shown promise in efficiently producing near-optimal results with respect to shape similarity as well as compliance with negligible run-time cost [34][35][36][37][38][39][40][41]. Theory-guided machine learning methods use domain-specific theories to establish the mapping between the design variables and the external boundary conditions [42][43][44].…”

Toward Hybrid Teams: A Platform to Understand Human-Computer Collaboration During the Design of Complex Engineered Systems

“…Two design agents were constructed on top of Energy3D to scaffold divergent and convergent design processes, respectively, both of which were powered by genetic algorithms (Schimpf et al, 2018). Furthermore, generative design agents have been developed to produce design alternatives utilizing Generative Adversarial Networks (GANs) (Dering and Tucker, 2017), Recurrent Neural Networks (RNNs) (Stump et al, 2019), and convolutional networks (Dosovitskiy et al, 2017;Raina et al, 2019). A generative design module has also been included in Siemens NX (Haubrock and Bevan, 2017).…”

A sparsity preserving genetic algorithm for extracting diverse functional 3D designs from deep generative neural networks

“…Oh et al (2018) combine GNNs and topology optimization to optimize the shape of generated images of wheels according to their compliance. Raina et al (2019) propose a deep learning method to learn to sequentially draw 2D truss designs from a data set of human pen strokes. This approach allows the learning agent to dynamically participate in the design process with a human designer, and is not mutually exclusive with the proposed method that is aimed at providing a human designer with a population of conceptual candidate designs as potential starting points.…”