Machine Learning Geometry Models for Multi-Platform Space Design
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for creating, updating, and modifying space configurations are labor-intensive and inefficient, often resulting in delayed processing times and increased processing loads due to the complexity of variables and diverse documentation methods.
Innovation Solution
A computing platform utilizing machine learning techniques and generative design algorithms to generate and rank space models based on input parameters, using a tiered approach to reduce processing time and computational bandwidth by solving for blocks, settings, and furniture in sequential stages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional manual processes are used to create and modify space configurations, then design flexibility and customization are maintained, but processing time and labor intensity increase significantly
Solution Approach 1:
The system enables self-service automated generation of space models by the computing platform itself, using machine learning algorithms to autonomously create, score, and rank multiple space model options based on input parameters without requiring manual intervention for each design iteration
Solution Approach 2:
The system changes parameters by automatically varying design variables such as furniture placement, space utilization, and layout configurations to generate multiple unique space models that satisfy the same design constraints, thereby maintaining design flexibility while reducing manual processing time
2Adaptability or versatility
If conventional systems generate multiple space model variants with different furniture and settings, then design options increase, but computational load and processing time increase exponentially
Solution Approach 1:
The system segments the space model generation process into distinct hierarchical levels: block models (overall space organization), settings models (furniture arrangements within blocks), and individual furniture models. This segmentation allows the system to generate and evaluate multiple design options at each level independently, increasing adaptability while managing computational complexity through modular processing
Solution Approach 2:
The system applies partial action by generating a large number of space model variants beyond what a single manual designer would typically create, then using automated scoring mechanisms to identify the most promising options. This excessive generation of variants is made computationally feasible through the segmented approach and automated evaluation, providing superior design versatility without proportional increases in manageable complexity
3Adaptability or versatility
If multiple documentation formats are supported for different design tools, then interoperability improves, but data processing complexity and bandwidth consumption increase
Solution Approach 1:
The computing platform implements multi-functionality by generating space models in multiple documentation formats (JSON, CSV, XML, and tool-specific formats for Autodesk Revit, Autodesk AutoCAD, and Graphisoft ArchiCAD) simultaneously. This universal format support improves interoperability with different design tools while the system manages data processing efficiency through standardized internal representations that are selectively exported to different formats based on user needs
Data Source
AI summary
Aspects of the disclosure relate to geometry model generation. A computing platform may receive a plurality of drawing models corresponding to different space designs. The computing platform may identify a plurality of design parameters associated with each drawing model of the plurality of drawing models corresponding to the different space designs. The computing platform may train a machine learning engine based on the plurality of drawing models corresponding to the different space designs and the plurality of design parameters associated with each drawing model of the plurality of drawing models corresponding to the different space designs, which may produce at least one geometry model corresponding to the plurality of drawing models. The computing platform may store, in a database storing one or more additional geometry models, the at least one geometry model corresponding to the plurality of drawing models.