Adaptive Robotic Assembly Using ML-Based Re-Grasp Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robotic assembly systems, known as fixed automation, are not adaptable to dynamic manufacturing environments and require significant reprogramming for changes in assembly processes or parts, making them inefficient in handling variations in product assembly.
Innovation Solution
A computer-implemented method using machine learning models to predict grasp proposals and pose estimates based on sensor data, allowing a robotic system to autonomously adjust its grasping and movement strategies for assembly tasks in a dynamic environment, enabling easier configuration for new or modified assembly processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If fixed automation is used to program robotic systems for repeated assembly actions, then manufacturing precision and reliability are improved, but adaptability and ease of reconfiguration deteriorate
Solution Approach 1:
The robotic system transitions from static fixed automation to dynamic adaptive automation by incorporating machine learning models that continuously learn from sensor data. The system dynamically adjusts grasp proposals and pose estimates based on real-time environmental changes, part variations, and assembly conditions, enabling it to adapt to dynamic manufacturing environments while maintaining reliable assembly operations.
Solution Approach 2:
The system changes operational parameters dynamically by using machine learning models to predict grasp proposals and pose estimates based on sensor data. Instead of fixed programmed parameters, the system adapts parameters such as grasp force, robot positioning, and assembly sequence based on learned patterns from environmental feedback, enabling flexibility while maintaining precision.
2Manufacturing precision
If fixed automation programming is used for robotic assembly, then manufacturing precision is improved, but reprogramming time and complexity increase
Solution Approach 1:
The robotic system performs self-learning and selfconfiguration through machine learning models that automatically adapt to new assembly tasks. Instead of requiring external reprogramming, the system learns from sensor data during operation, automatically adjusting its behavior to maintain precision when assembling different parts or modifying assembly processes, thereby eliminating time-consuming reprogramming.
Solution Approach 2:
The system performs preliminary learning and adaptation during initial operation phases by collecting sensor data and training machine learning models. This preliminary action enables the system to pre-adapt to new assembly configurations or part variations before actual production, reducing the time needed for reprogramming when changes are required.
3Stability of the object's composition
If fixed automation is implemented for robotic assembly, then operational stability is improved, but ease of operation and flexibility deteriorate
Solution Approach 1:
The robotic system incorporates continuous feedback loops where sensor data from the assembly environment is fed into machine learning models that predict grasp proposals and pose estimates. This feedback mechanism enables the system to automatically adjust its operations based on real-time conditions, making it easier to handle variations in parts, conveyor positions, or assembly requirements while maintaining stable and reliable operation through learned patterns.
Data Source
AI summary
Techniques are disclosed for controlling robotic systems to perform assembly tasks. In some embodiments, a robot control application receives sensor data associated with one or more parts. The robot control application applies a grasp perception model to predict one or more grasp proposals indicating regions of the one or more parts that a robotic system can grasp. The robot control application causes the robotic system to grasp one of the parts based on a corresponding grasp proposal. If the pose of the grasped part needs to be changed in order to assemble the part with one or more other parts, the robot control application determines movements of the robotic system required to re-grasp the part in a different pose. In addition, the robot control application determines movements of the robot system for assembling the part with the one or more other parts based on results of a motion planning technique.