3D Assembly Asset Generation With Contact Surfaces and Clearance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robotic assembly techniques struggle to train robots to assemble a range of parts with diverse geometries reliably, as existing datasets are limited and often include interpenetrating assets, limiting their applicability in both simulation and real-world environments.
Innovation Solution
A three-stage pipeline is employed to automate the generation of paired parts, involving contact surface extraction, shape completion, and clearance specification, enabling the creation of a diverse set of assembly asset pairs that can be used to train and evaluate robots in simulated and real-world environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If manual processes are used to generate assembly datasets, then the quality and physical feasibility of asset pairs can be controlled, but the quantity and diversity of assembly problems are limited
Solution Approach 1:
The patent replaces manual mechanical curation processes with an automated neural network-based system. The neural network automatically generates and filters assembly asset pairs, substituting human manual selection with AI-driven generation that can process vast numbers of combinations efficiently while maintaining physical feasibility through learned constraints.
Solution Approach 2:
The system employs self-service mechanisms where the neural network automatically generates, evaluates, and filters assembly asset pairs without requiring manual intervention. The automated pipeline independently handles the entire process of creating diverse and physically feasible datasets, eliminating the need for human curators to manually verify each asset pair.
2Adaptability or versatility
If existing datasets are used for training, then the training process can proceed with available data, but the robot cannot reliably assemble parts with diverse geometries
Solution Approach 1:
The patent changes the parameters of the training data by using a neural network to generate asset pairs with varied geometric properties. The system can control parameters such as shape complexity, size variations, and geometric diversity to create training datasets that enable robots to handle a broader range of part geometries while maintaining data quality and physical feasibility.
3Ease of manufacture
If conventional techniques generate assembly datasets, then the generation process is simple, but the datasets include interpenetrating assets that cannot be physically assembled
Solution Approach 1:
The system incorporates feedback mechanisms where the neural network continuously evaluates generated asset pairs against physical feasibility constraints. The network receives feedback about interpenetration and clearance issues and automatically adjusts generation parameters or filters out invalid pairs, ensuring that only physically assembleable asset pairs are included in the final dataset while maintaining ease of automated generation.
Data Source
AI summary
In various examples, a three-stage pipeline is used to automate the generation of paired parts (or components) in assemblies. The pipeline includes a first contact surface extraction stage, in which a set of contact surfaces is extracted from a first part based on attributes identified by a vision language model (VLM) and/or another type of machine learning model from a visual and/or another representation of the first part. The pipeline also includes a shape completion stage, in which the contact surfaces are used to condition the operation of a diffusion model and/or another type of three-dimensional (3D) generative model in generating a shape for a second part that is complementary to the first part. The pipeline further includes a clearance specification stage, in which the shape of a given part is updated to meet a minimum clearance distance from the other part.