Robot Posture Estimation with Virtual Markers Under Occlusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately estimate the posture of a robot when one marker is obscured, and the exact posture of the robot's joints and links cannot be determined using conventional techniques, especially when external vision sensors are used.
Innovation Solution
A method and system that utilize at least one first marker attached to the robot's end-effector and multiple second markers, estimating their positional relationship to derive the robot's posture even when the first marker is not identified, using image processing and machine learning to determine the robot's posture based on the relative positional information of the markers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If external vision sensors are used to detect robot defects, then defect detection capability is improved, but the ability to estimate robot posture when markers are obscured deteriorates
Solution Approach 1:
The patent introduces a virtual marker as an intermediary element to solve the problem of marker obscuration. When the physical first marker is obscured or not detected, the system creates a virtual marker at the expected position based on the detected second markers and pre-stored positional relationships. This virtual marker serves as a mediator to maintain continuous posture estimation without interruption, thereby resolving the contradiction between defect detection capability and posture estimation accuracy.
Solution Approach 2:
The patent creates a virtual copy of the first marker (virtual marker) based on the detected second markers and pre-stored spatial relationships. This copy allows the system to continue estimating robot posture even when the original marker is obscured. The virtual marker is generated by calculating the expected position and orientation of the first marker relative to the detected second markers, enabling continuous monitoring without sacrificing measurement precision.
2Device complexity
If only one marker is attached to the robot, then device complexity is reduced, but the ability to estimate posture when the marker is obscured deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-storing the positional relationships between the first marker and multiple second markers before robot operation. These spatial relationships are captured and saved in advance, allowing the system to quickly generate virtual markers when needed without requiring additional real-time computation or hardware complexity. This pre-prepared information enables reliable posture estimation even when markers are obscured.
3Measurement precision
If multiple cameras are used to recognize marker positions, then measurement precision is improved, but device complexity and installation difficulty increase
Solution Approach 1:
The patent makes the vision system universal by enabling a single camera to perform multiple functions: detecting physical markers, identifying their spatial relationships, and generating virtual markers. This multi-functional approach eliminates the need for multiple cameras while maintaining measurement precision. The system can adaptively switch between detecting real markers and generating virtual markers based on detection results, reducing hardware complexity while preserving accuracy.
Data Source
AI summary
A method of estimating a posture of a robot is disclosed. The posture estimation method of the robot according to the present disclosure includes: generating a first marker and at least one second marker and attaching them to the robot; collecting an image including the first marker and at least one of the plurality of second markers; estimating position information and rotation information of the first marker and each of the second markers; deriving the position information and the rotation information of the first marker depending on a relative positional relationship between the first marker and each of the second markers when the first marker is not identified; and estimating a posture of the robot based on the position information and the rotation information of the first marker.