Robot Grasp Planning Using 2D Vision and Grabbing Templates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for a robot to grasp a 3D object using a 2D camera are inefficient due to complex environment detection issues, high costs associated with 3D camera-based techniques, and redundancy in six parameter representations, leading to increased computational burden and lack of flexibility in grabbing points and positions.
Innovation Solution
A method involving a visual sensor to determine the current position and attitude of a robot relative to a 3D object, using a grabbing template with specified and reference positions and attitudes, optimized by hand-eye calibration, which reduces computational complexity and eliminates the need for expensive 3D cameras.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 3D camera-based techniques are used to estimate position and attitude, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent creates a virtual model (copy) of the 3D object and simulates multiple viewing angles to generate a virtual image library. This virtual copy replaces the need for expensive 3D cameras, allowing the system to estimate position and attitude by matching real 2D camera images against the pre-generated virtual images, thereby achieving accurate measurement with simpler equipment
Solution Approach 2:
The patent replaces expensive 3D camera hardware with a computational approach using standard 2D cameras and pre-generated virtual image data. The virtual image library acts as a disposable computational resource that can be generated once and reused multiple times, eliminating the need for continuous expensive hardware investment
2Loss of information
If six parameter representation (x, y, z, roll, pitch, yaw) is used to indicate position and attitude, then completeness of information is improved, but computational burden increases
Solution Approach 1:
The patent extracts only the essential position and attitude parameters needed for grabbing operations from the full six-parameter representation. By focusing on the specific parameters relevant to the grabbing task rather than all six parameters, the system reduces computational burden while maintaining the information completeness required for the specific application
Solution Approach 2:
The patent applies partial action by using a simplified parameter representation that covers only the necessary degrees of freedom for grabbing operations. Rather than processing all six parameters equally, the system selectively processes the subset of parameters that are most critical for the grabbing task, reducing overall computational complexity
3Productivity
If template matching or deep learning is used to estimate position and attitude directly, then productivity is improved, but adaptability to different environments deteriorates
Solution Approach 1:
The patent performs preliminary action by pre-generating a comprehensive library of virtual images covering multiple viewing angles, lighting conditions, and object orientations before actual grabbing operations. This pre-computed virtual image library enables rapid matching during operation while maintaining adaptability to different environments, as the library can be generated to include various environmental conditions
Solution Approach 2:
The patent implements dynamics by creating a flexible virtual image library that can be dynamically adjusted to match different environmental conditions. The system can generate new virtual images on-demand for unseen environments, allowing the matching process to adapt to varying lighting, angles, and object positions while maintaining fast operation speeds
Data Source
AI summary
Various embodiments include a method for a robot to grab a 3D object. The method may include: determining a current position and attitude of a visual sensor of the robot relative to the 3D object; acquiring a grabbing template of the 3D object, the grabbing template comprising a specified grabbing position and attitude of the visual sensor relative to the 3D object; judging whether the grabbing template further comprises at least one reference grabbing position and attitude of the visual sensor relative to the 3D object, wherein the reference grabbing position and attitude is generated on the basis of the specified grabbing position and attitude; and based on a judgment result, using the grabbing template and the current position and attitude to generate a grabbing position and attitude of the robot.