Robot Gripper Hold Position Mapping Without Extensive Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing learning devices require a large amount of teacher data to accurately determine the holding position of an object in line with human intention, increasing workload and cost.
Innovation Solution
A holding position determination device and method that uses a control unit to acquire an end effector model and a rule map based on images of the object, determining the holding position by projecting the end effector model onto various maps to evaluate appropriateness and calculate a coincidence degree, thereby simplifying the determination process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a learning device is used to determine holding position, then accuracy in determining holding position in line with human intention is improved, but the amount of teacher data required increases, leading to increased workload and cost
Solution Approach 1:
The patent uses a rule map that copies human decision-making logic for holding position determination. Instead of requiring extensive teacher data to train a learning device, the system creates a rule map that represents human intention and reasoning processes, allowing the robot to determine holding positions by referencing this mapped knowledge rather than learning from large datasets
Solution Approach 2:
The rule map serves as an intermediary between human intention and robot execution. Rather than directly training a learning device with numerous examples, the patent introduces a rule map that translates human reasoning into a structured format that the robot can follow, eliminating the need for extensive teacher data while maintaining accuracy
2Measurement precision
If a learning device requires extensive teacher data, then holding position determination accuracy is improved, but workload and cost increase
Solution Approach 1:
The system copies human reasoning processes into a rule map structure, allowing the robot to determine holding positions by following predefined rules that represent human intention. This approach maintains high accuracy while significantly reducing the complexity associated with collecting and processing extensive teacher data
3Measurement precision
If traditional learning methods are used, then holding position accuracy is improved, but adaptability to environmental changes decreases
Solution Approach 1:
The rule map is designed to be dynamically adjustable to environmental changes. When the environment changes (e.g., different objects, lighting conditions, or positions), the rule map can be updated with new rules or modified existing ones, allowing the system to adapt without retraining from scratch while maintaining determination accuracy
Data Source
AI summary
A holding position determination device includes a control unit that determines the position where an end effector adapted to hold a holding target contacts the holding target as a holding position. The control unit acquires an end effector model that identifies an area where a holding portion of the end effector can exist. The control unit acquires a rule map that includes a map defining the position of the holding target to be held by the end effector based on an image of holding target obtained by photographing the holding target. The control unit determines the holding position based on the end effector model and the rule map.