3D Object Design via Latent Vector Optimization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

There is a need for an improved method for designing 3D modeled objects via user-interaction that leverages machine-learning techniques effectively.

Innovation Solution

The method utilizes a machine-learnt decoder to design 3D modeled objects by determining an optimal latent vector that minimizes an energy function penalizing non-respect of user-defined deformation constraints, thereby outputting a 3D modeled object that respects the user's design intent.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional CAD systems are used for designing 3D modeled objects, then design flexibility is maintained, but user interaction complexity and time consumption increase

Engineering Contradiction:
Improvedesign efficiencyVSAvoiduser interaction complexity
Core Design Contradiction:
ProductivityVSEase of operation

Solution Approach 1:

The patent introduces a machine learning model as an intermediary between the user and the complex CAD system. The user provides simple deformation constraints, and the ML model automatically generates the optimal 3D object configuration, eliminating the need for users to navigate complex CAD interfaces and perform manual modeling operations.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system enables self-service by allowing the 3D design system to automatically optimize object configurations based on user constraints. The machine learning model independently performs the complex design optimization tasks without requiring continuous user intervention or adjustment of multiple parameters.

Inventive Principle:
Principle #25Self-service

2Extent of automation

If machine learning techniques are integrated into 3D design systems, then design automation improves, but system complexity increases

Engineering Contradiction:
Improvedesign automationVSAvoidsystem complexity
Core Design Contradiction:
Extent of automationVSDevice complexity

Solution Approach 1:

The patent uses a pre-trained machine learning model that has been trained on a large dataset of 3D objects. This pre-trained model can be directly applied to new design tasks without requiring retraining, simplifying the system architecture and reducing the complexity of integrating ML capabilities into the CAD system.

Inventive Principle:
Principle #26Copying

3Manufacturing precision

If user-defined deformation constraints are applied, then design precision improves, but computational requirements increase

Engineering Contradiction:
Improvedesign precisionVSAvoidcomputational energy
Core Design Contradiction:
Manufacturing precisionVSUse of energy by moving object

Solution Approach 1:

The machine learning model is pre-trained on a comprehensive dataset of 3D objects and deformation patterns. This preliminary training enables the model to quickly infer optimal configurations from user constraints without requiring computationally intensive real-time calculations, thus reducing energy consumption during actual design operations.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3671660B1Designing a 3D modeled object via user-interaction
Publication Date: 2025.06.18 DASSAULT SYSTEMES SA
  • EP3671660B1 patent drawingFigure 1
  • EP3671660B1 patent drawingFigure 2
  • EP3671660B1 patent drawingFigure 3

AI summary

The invention notably relates to a computer-implemented method for designing a 3D modeled object via user-interaction. The method comprises providing the 3D modeled object and a machine-learnt decoder. The machine-learnt decoder is a differentiable function taking values in a latent space and outputting values in a 3D modeled object space. The method further comprises defining, by a user, a deformation constraint for a part of the 3D modeled object. The method further comprises determining an optimal vector. The optimal vector minimizes an energy. The energy explores latent vectors. The energy comprises a term which penalizes, for each explored latent vector, non-respect of the deformation constraint by the result of applying the decoder to the explored latent vector. The method further comprises applying the decoder to the optimal latent vector. This constitutes an improved method for designing a 3D modeled object via user-interaction.