Robot Vision Control Using Predicted Images for Self-Teaching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robot control systems require significant labor to create programs for robot operations and are inflexible in handling environmental changes, as they do not specify how the robot arm adjusts to target positions and directions.
Innovation Solution
A robot system incorporating an image acquisition part, an image prediction part, and an operation controller that uses machine learning to predict next images based on teaching images, allowing the robot to adjust its position and direction dynamically and adapt to environmental changes without pre-defined programs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a program describing robot operations is created in advance by a skilled person, then the robot can perform predetermined operations, but the labor and time required for program creation increases significantly
Solution Approach 1:
The robot system performs self-teaching by autonomously learning the relationship between movable part positions and imaging device images through automatic image acquisition and model construction, eliminating the need for manual program creation by skilled operators
Solution Approach 2:
The patent replaces manual programming operations with an automated machine learning system that constructs teaching image models and predicts next images algorithmically, substituting human expertise with computational processes
2Stability of the object's composition
If the robot operates based on pre-defined programs, then the operation sequence is determined, but the system cannot flexibly adapt to environmental changes
Solution Approach 1:
The system dynamically adapts to environmental changes by continuously acquiring current images, predicting next images based on updated teaching models, and adjusting movable part positions in real-time rather than following fixed pre-programmed sequences
Solution Approach 2:
The patent implements feedback control by comparing current images with predicted next images, calculating command values based on the difference, and adjusting the movable part positions iteratively to achieve accurate operation despite environmental variations
3Measurement precision
If the imaging device moves with the operation tool, then the robot can maintain visual tracking, but the system complexity increases
Solution Approach 1:
The patent merges the imaging device with the operation tool or positions it to move synchronously with the robot arm, creating an integrated visual-sensing system that maintains consistent observation of the workpiece throughout the operation sequence
Data Source
AI summary
A robot system includes a robot, an image acquisition part, an image prediction part, and an operation controller. The image acquisition part acquires a current image captured by a robot camera arranged to move with the end effector. The image prediction part predicts a next image to be captured by the robot camera based on a teaching image model and the current image. The teaching image model is constructed by a machine learning of teaching image which is predicted to capture by the robot camera while the movable part performs an adjustment operation. The operation controller calculates the command value for operating the movable part so that the image captured by the robot camera approaches the next image, and controls the movable part based on the command value.