Abstract:Generative systems have a significant potential to synthesize innovative design alternatives. Still, most of the common systems that have been adopted in design require the designer to explicitly define the specifications of the procedures and in some cases the design space. In contrast, a generative system could potentially learn both aspects through processing a database of existing solutions without the supervision of the designer. To explore this possibility, we review recent advancements of generative models in machine learning and current applications of learning techniques in design. Then, we describe the development of a data-driven generative system titled DeepCloud. It combines an autoencoder architecture for point clouds with a web-based interface and analog input devices to provide an intuitive experience for data-driven generation of design alternatives. We delineate the implementation of two prototypes of DeepCloud, their contributions, and potentials for generative design.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: **How to use data - driven generative models to automatically generate novel design proposals in the design field without designers explicitly defining the design process and design space**. Specifically, traditional generative systems (Generative Systems) usually require designers to clearly specify the parameters in the design process and the design space. This limits the flexibility and innovativeness of the system. In contrast, data - driven generative models based on machine learning can automatically learn the design space and the generation process by processing the existing database of design solutions, thereby reducing the need for designer supervision and increasing the diversity and innovativeness of the design. ### Core problems of the paper 1. **Limitations of the traditional design process**: - In the traditional design process, designers need to clearly define the design problem and explore solutions. - Although the generative system can automatically generate design proposals, it still requires designers to clearly specify the generation process and the design space. 2. **Potential of data - driven generative models**: - Data - driven generative models can learn from a large number of existing design cases and automatically discover the design space and generation rules. - This method can reduce the dependence on designers and generate more diverse and innovative design proposals at the same time. 3. **Deficiencies in existing research**: - Currently, most design - generation systems still rely on designers' input and parameter settings. - Although some generative models based on machine learning have been applied in the design field, their application scope is limited, and they lack interactivity and performance - evaluation standards. ### Goals of the paper To make up for the above - mentioned deficiencies, this paper proposes a data - driven generative system named **DeepCloud**, aiming to achieve the following goals: - **Data - driven design space**: Explore techniques that can learn the design space from a large - example database instead of relying on designers' manual parameterization. - **Machine - learning techniques**: Use the latest deep - generative models (such as auto - encoders, variational auto - encoders, etc.) for design generation. - **Representation methods**: Use general and flexible geometric representation methods to adapt to the design modeling of multiple problem domains. - **Operation space**: Explore the ability to embed meaningful geometric transformations in high - dimensional space, not just the visualization of low - dimensional space. Through these goals, DeepCloud attempts to provide designers with a more intuitive and flexible tool for generating and exploring novel design proposals.

DeepCloud. The Application of a Data-driven, Generative Model in Design

Generative Creativity: Adversarial Learning For Bionic Design

Generative Artificial Intelligence Enhances Sudden Moments of Inspiration among Novice Designers

Neurocognition-inspired Design with Machine Learning

Deep Generative Design: Integration of Topology Optimization and Generative Models

Generative Deep Learning in Architectural Design

Deep Generative Models in Engineering Design: A Review

Towards Goal, Feasibility, and Diversity-Oriented Deep Generative Models in Design

Exploring the Potentials and Challenges of Deep Generative Models in Product Design Conception

Human-in-the-loop design with machine learning

Generative AI for 3-D Point Clouds

Product Design Using Generative Adversarial Network: Incorporating Consumer Preference and External Data

DesignFusion: Integrating Generative Models for Conceptual Design Enrichment

Generative Design: An Overview and Its Relationship to Artificial Intelligence

A Hybrid Prototype Method Combining Physical Models and Generative Artificial Intelligence to Support Creativity in Conceptual Design

Deep learning framework for material design space exploration using active transfer learning and data augmentation

Deep Concept Identification for Generative Design

Generative Design: Reframing the Role of the Designer in Early-Stage Design Process

Data-driven generative design for mass customization: A case study

From Cloud to Edge: Rethinking Generative AI for Low-Resource Design Challenges

Inspired by AI? A Novel Generative AI System To Assist Conceptual Automotive Design