Robot Cell 3D Calibration Using Mobile Sensing and Robot Geometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining and calibrating a 3D representation of a robot cell are inefficient and prone to collisions, requiring additional sensors or robot motion, which can result in incomplete data capture and potential collisions with objects.
Innovation Solution
A method using a mobile sensor to capture cell data from multiple viewpoints, guided by a graphical user interface, and calibrating the 3D representation using known robot geometry to ensure accurate location, orientation, and scale, with iterative data capture until quality thresholds are met.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensors are mounted at known locations with calibration markers, then the location and scale of captured data can be calibrated, but additional expensive hardware components are required and the solution is not adaptable when the environment changes
Solution Approach 1:
The patent uses the robot's own known geometric structure as a reference copy for calibration, eliminating the need for external calibration markers. The robot's geometry serves as a virtual reference that can be matched against the captured point cloud data to determine transformation parameters.
Solution Approach 2:
The robot performs self-calibration by using its own known geometry as the reference. The system leverages the pre-known robot structure to automatically calculate transformation parameters without requiring external calibration equipment or manual intervention.
2Ease of operation
If sensors are mounted on the robot, then environmental data is captured with respect to the robot location, but the robot must move to scan the complete robot cell which may result in collisions with objects
Solution Approach 1:
The patent performs preliminary planning of the robot's scanning path to ensure complete coverage of the robot cell while avoiding collisions. The system calculates optimal viewpoints and motion trajectories before execution, preventing collisions with undetected objects.
Solution Approach 2:
The system uses feedback from the captured point cloud data to adjust the scanning process. After each scanning position, the system evaluates whether complete coverage is achieved and determines whether additional viewpoints are needed, optimizing the scanning process dynamically.
3Productivity
If SLAM is used to generate environment representation during robot motion, then a 3D representation can be created, but algorithms are required to determine robot location for calibration and the process is time-consuming
Solution Approach 1:
The patent replaces complex SLAM algorithms with a simpler geometric matching approach. Instead of using probabilistic localization methods, the system directly matches the robot's known geometry with the captured point cloud to determine transformation parameters, significantly reducing computation time.
Solution Approach 2:
The system changes the calibration approach from iterative SLAM optimization to direct geometric parameter matching. By using the robot's pre-known geometry as a reference, the transformation parameters can be calculated directly without iterative optimization, reducing calibration time.
4Loss of information
If additional sensors or robot motion is used to capture data from further areas, then complete coverage can be achieved, but device complexity increases or collision risk increases
Solution Approach 1:
The patent uses dynamic planning of the robot's scanning positions and orientations to achieve complete coverage. The system adaptively determines optimal viewpoints based on the current scan state and evaluates coverage completeness, allowing a single sensor to capture the entire environment without additional hardware.
Data Source
AI summary
A method for determining and calibrating a 3D representation of a robot cell includes a) obtaining cell data that includes environment data from a mobile sensor, wherein the environment data comprises a set of images of the robot cell from a first viewpoint; b) determining a 3D representation of the robot cell using the cell data; c) calibrating the 3D representation using a known geometry of the industrial robot; d) visualizing the 3D representation at the calibrated location and orientation in the robot cell in simulation software; e) determining the quality of the 3D representation; and f) repeating steps a)-e) when the quality of the 3D representation is smaller than a predetermined threshold.