Robot Gripping with Machine Learning for Bulk Workpiece Picking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing gripping systems for robots struggle to efficiently grip bulk-loaded workpieces of varying shapes and sizes due to limitations in recognizing and adapting to the three-dimensional position and posture of individual workpieces, often relying on distance sensors and 3D CAD models which are not efficient for diverse workpiece arrangements.
Innovation Solution
A gripping system that incorporates a robot hand, an image sensor, and a machine learning-based model to acquire and process image information and hand position data, enabling the derivation and execution of operation commands for precise gripping, independent of traditional distance sensors and 3D CAD models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If distance sensors and 3D CAD models are used to recognize workpiece position and posture, then measurement capability is provided, but the system complexity and inefficiency increase for diverse workpiece arrangements
Solution Approach 1:
The patent replaces the mechanical sensing system (distance sensors physically measuring workpiece positions) with a vision-based system using image sensors and machine learning. The image sensor captures images of bulk-loaded workpieces, and a machine learning model automatically recognizes three-dimensional positions and postures from these two-dimensional images, eliminating the need for complex mechanical measurement systems while handling diverse workpiece arrangements efficiently
Solution Approach 2:
The patent creates a virtual model (copy) of the workpiece arrangement by generating a three-dimensional model from two-dimensional image data through machine learning. This virtual model replicates the physical arrangement of bulk-loaded workpieces, allowing the system to recognize positions and postures without direct physical measurement, thereby reducing system complexity while maintaining measurement precision
2Productivity
If traditional gripping systems are used for bulk-loaded workpieces, then gripping operation is performed, but the gripping success probability decreases due to inability to adapt to varying shapes and sizes
Solution Approach 1:
The patent implements a dynamic gripping system where the robot's hand position and posture are continuously adjusted based on real-time image recognition results. The machine learning model adapts to varying workpiece shapes and sizes by learning from training data, enabling the system to dynamically optimize gripping parameters for each specific workpiece configuration, thereby maintaining high gripping success probability across diverse arrangements
Solution Approach 2:
The patent establishes a feedback loop where image sensors continuously monitor workpiece positions and postures, the machine learning model processes this visual information to determine optimal gripping parameters, and the robot adjusts its hand position accordingly. This closed-loop feedback system enables real-time adaptation to varying workpiece arrangements, improving both gripping reliability and overall productivity
3Measurement precision
If distance sensors are fixedly disposed on workpieces, then measurement is enabled, but the ease of operation and adaptability to diverse arrangements deteriorate
Solution Approach 1:
The patent employs a universal image sensing system that can measure positions and postures of various workpiece types and arrangements through a single imaging device. Unlike fixed distance sensors that require specific mounting configurations, the image sensor captures visual information from bulk-loaded workpieces in any arrangement, and the machine learning model processes this data universally, greatly improving ease of operation and adaptability
4Measurement precision
If 3D CAD models are collated with measurement results, then position recognition is achieved, but the loss of time increases due to complex processing requirements
Solution Approach 1:
The patent replaces the time-consuming process of collating measurement results with 3D CAD models by using a machine learning model that directly predicts three-dimensional positions and postures from two-dimensional images. This substitution eliminates the iterative matching process between sensor data and CAD models, significantly reducing processing time while maintaining accurate position recognition for bulk-loaded workpieces
Data Source
AI summary
A gripping system includes a hand that grips a workpiece, a robot that supports the hand and changes at least one of a position and a posture of the hand, and an image sensor that acquires image information from a viewpoint interlocked with at least one of the position and the posture of the hand. Additionally, the gripping system includes a construction module that constructs a model by machine learning based on collection data. The model corresponds to at least a part of a process of specifying an operation command of the robot based on the image information acquired by the image sensor and hand position information representing at least one of the position and the posture of the hand. An operation module executes the operation command of the robot based on the image information, the hand position information, and the model, and a robot control module operates the robot based on the operation command of the robot operated by the operation module.