Robot Pose Estimation Using Descriptor Image Feature Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robotic learning-from-demonstration (LfD) approaches face challenges in reliable pose estimation of objects, particularly due to the unreliability of RGB-image based methods.
Innovation Solution
A method for controlling robots that involves selecting feature points based on an objective function rewarding visibility and spatial separation, training a robot control model using these points, and estimating the object pose for accurate robot control, even in scenarios with non-real 3D objects, by using a dense object net for descriptor image generation and Euclidean coordinate determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If RGB-image based pose estimation techniques are used, then the implementation is simple, but the reliability of pose estimation deteriorates
Solution Approach 1:
The patent replaces traditional RGB-image based pose estimation with a method using descriptor images and feature point detection. Instead of relying on color and texture information from RGB images, the system uses binary descriptors (FREAK, ORB, BRISK) that are more robust to lighting and appearance variations, thereby improving reliability while maintaining implementation feasibility
Solution Approach 2:
The patent changes the parameter space for pose estimation by transitioning from RGB color space to descriptor space. The objective function optimizes feature point selection based on visibility and spatial distribution in descriptor images rather than relying on RGB image characteristics, making the estimation more reliable across different lighting conditions
2Reliability
If feature points are selected to be highly visible in descriptor images, then pose estimation reliability improves, but the complexity of feature point selection increases
Solution Approach 1:
The system performs self-selection of feature points by automatically optimizing the objective function that evaluates visibility and spatial distribution. The algorithm autonomously identifies the most informative feature points without requiring manual intervention or complex configuration, balancing reliability improvement with implementation simplicity
Solution Approach 2:
The patent uses descriptor images as a simplified representation or copy of the actual object appearance. These binary descriptors capture essential geometric features without the complexity of full RGB images, enabling reliable feature point selection while reducing computational burden
3Manufacturing precision
If the robot control model uses accurate pose estimation, then the control precision improves, but the computational time increases
Solution Approach 1:
The patent extracts only the essential feature points from the descriptor images that are most informative for pose estimation. By selecting a small subset of key feature points based on the objective function rather than processing all possible points, the system achieves high control precision while reducing computational time
Solution Approach 2:
The system uses partial action by estimating pose from a selected subset of feature points rather than using all available information. This selective approach provides sufficient precision for control while significantly reducing the computational burden compared to exhaustive methods
Data Source
AI summary
A method for controlling a robot. The method includes performing demonstrations and descriptor images for the demonstrations from a point of view of the robot of the object; selecting a set of feature points, wherein the feature points are selected by searching an optimum of an objective function which rewards selected feature points being visible in the descriptor images; training a robot control model using the demonstrations and controlling the robot for a control scene with the object by determining a descriptor image of the object, locating the selected set of feature points in the descriptor image of the object; determining Euclidean coordinates of the located feature points; estimating a pose from the determined Euclidean coordinates; and controlling the robot to handle the object by means of the robot control model with the estimated pose.