Robot Assembly Control Using Reversed Disassembly Trajectories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional robotic assembly systems require customized fixtures and predefined waypoints, limiting their applicability in high-mixture settings where various part shapes, sizes, and orientations are involved, necessitating extensive reconfiguration.
Innovation Solution
A computer-implemented method for training a robot control model by generating disassembly trajectories, reversing them to create assembly trajectories, and using machine learning models to adapt robotic movements based on real-time part observations, eliminating the need for customized fixtures and predefined tooling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional robotic assembly systems use customized fixtures and predefined waypoints, then manufacturing precision and reliability are improved, but adaptability to high-mixture settings deteriorates
Solution Approach 1:
The patent replaces mechanical fixtures and predefined mechanical waypoints with a machine learning-based control system. The ML model learns assembly trajectories from disassembly demonstrations and generalizes to different part configurations, eliminating the need for physical fixtures and predefined tooling for each part variation.
Solution Approach 2:
The system changes the control parameters from fixed predefined waypoints to dynamically generated trajectories based on ML predictions. The robot controller adjusts motion parameters in real-time based on sensor feedback and ML model outputs, enabling adaptation to different part geometries without reconfiguring mechanical fixtures.
2Manufacturing precision
If robotic assembly systems are configured for specific part shapes and sizes, then manufacturing precision is improved, but device complexity increases due to reconfiguration requirements
Solution Approach 1:
The patent creates a universal robotic assembly system that can handle multiple part types and configurations using a single ML-trained policy. The system performs multiple assembly functions across different product models without requiring reconfiguration of fixtures or tooling, as the ML model generalizes from training data to handle variations.
Solution Approach 2:
The system uses disassembly demonstrations as training data to create a digital model of assembly trajectories. By copying and reversing the disassembly motion patterns learned from demonstrations, the system generates assembly trajectories for various part configurations without physical reconfiguration.
3Manufacturing precision
If customized fixtures and tooling are used for each assembly task, then manufacturing precision is improved, but loss of time increases due to reconfiguration downtime
Solution Approach 1:
The system performs preliminary learning by training the ML model on disassembly demonstrations before actual assembly operations. This preliminary training phase creates a generalized policy that enables rapid adaptation to different part types during production, eliminating the need for time-consuming reconfiguration of fixtures and tooling between tasks.
Data Source
AI summary
The disclosed method for training a robot control model includes generating, using one or more simulations, a plurality of disassembly trajectories along which a first part is disassembled from a second part; reversing the plurality of disassembly trajectories to generate a plurality of reversed disassembly trajectories; and performing, based on the plurality of reversed disassembly trajectories, one or more operations to train an untrained machine learning model to generate a trained machine learning model, wherein the trained machine learning model is trained to control a robot to assemble the first part and the second part.