Dual-Camera Robot Gripping Position Correction for Unknown Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robot gripping systems face challenges in accurately gripping objects due to position errors from image analysis and mechanical defects, and struggle with unknown objects lacking shape data registration.
Innovation Solution
A robot control apparatus and method that utilizes dual cameras (overhead and hand-mounted) to generate encompassing boxes for objects, calculating a corrected gripping position based on relative positions from both cameras' views, ensuring accurate and stable gripping.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a hand camera is mounted on the hand unit to capture images of the object gripping process, then the object recognition accuracy is improved, but the range of grippable objects is limited to those in the vicinity of the hand unit
Solution Approach 1:
The patent combines the overhead camera and hand camera into a coordinated dual-camera system. The overhead camera captures the entire workspace to determine gross hand movement, while the hand camera captures detailed object views for precise recognition. By merging the functional roles of both cameras, the system achieves both wide coverage and high precision object recognition.
Solution Approach 2:
The patent introduces a hierarchical dimension to the vision system by assigning different observational perspectives to each camera. The overhead camera provides a top-down macro view for trajectory planning, while the hand camera provides a close-up micro view for detailed object recognition. This dimensional differentiation allows the system to simultaneously achieve wide coverage and high precision.
2Productivity
If shape data of objects is stored in advance for recognition, then the object recognition speed is improved, but the system cannot handle unknown objects without registered shape data
Solution Approach 1:
The system performs preliminary actions by pre-calculating and storing encompassing boxes for known objects based on their shape data. When an object is recognized, its pre-computed encompassing box is retrieved and used immediately, enabling fast recognition speeds. This preliminary preparation allows the system to efficiently handle known objects while maintaining the flexibility to process unknown objects through real-time camera coordination.
3Measurement precision
If the overhead camera captures both the object and hand unit, then the overall positioning is improved, but the gripping range is limited to objects near the hand unit
Solution Approach 1:
The patent segments the vision task into two distinct functions performed by separate cameras. The overhead camera is dedicated to capturing the hand unit position and trajectory, while the hand camera is dedicated to capturing object details and confirming gripping positions. This functional segmentation allows each camera to optimize its performance for its specific task, overcoming the limitations of a single camera trying to perform both functions.
Data Source
AI summary
A system and method generate an overhead camera-based encompassing box that encompasses an object to be gripped included in an image captured by an overhead camera mounted on a robot and a hand camera-based encompassing box that encompasses the object to be gripped included in an image captured by a hand camera mounted on the robot. Moreover, a relative position of a target gripping position of the object to be gripped with respect to the overhead camera-based encompassing box in the image captured by the overhead camera is calculated, the target gripping position with respect to the hand camera-based encompassing box in the image captured by the hand camera is calculated on the basis of the calculated relative position, and the calculated position is set as a corrected target gripping position of the object to be gripped included in the image captured by the hand camera.