Robot Arm Control Using Provisional Operation and User Corrections
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional robot control methods require significant time and effort from operators to program and adjust robot operations, especially when using machine learning for autonomous operation, leading to a high burden on operators for data collection and training.
Innovation Solution
A robot control device and method that utilizes machine learning to predict output data from input data, including the state of the robot and its surroundings, allowing for provisional operation, data collection, trained model building, and modified control, reducing operator burden and enabling high-quality training data collection even from non-skilled operators.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If machine learning is used to control robot operation, then autonomous operation capability is improved, but the burden on the operator for data collection increases
Solution Approach 1:
The patent segments the training data collection process into two distinct phases: provisional operation data collection (where the robot performs basic operations with human guidance) and modification work data collection (where the robot learns from correcting its own mistakes). This segmentation allows the operator to focus on providing correction data rather than collecting all training data, significantly reducing the operator burden while maintaining high autonomous operation capability.
Solution Approach 2:
The system performs preliminary action by first collecting provisional operation data where the robot attempts operations with basic guidance. This preliminary data collection establishes a baseline model that can then be refined through modification work data. The operator only needs to provide correction data for the robot's mistakes, rather than collecting all training data from scratch, thus reducing burden while achieving high autonomy.
2Measurement precision
If a large amount of training data is collected for machine learning, then model accuracy is improved, but the time required for data collection increases
Solution Approach 1:
The patent implements feedback by having the robot perform provisional operations and then having the operator or system provide correction data for mistakes. This feedback loop allows the model to learn from errors efficiently. The system collects modification work data where the robot learns what corrections are needed, achieving high model accuracy without requiring exhaustive data collection, thus reducing data collection time while maintaining precision.
Solution Approach 2:
The robot performs self-service by autonomously executing provisional operations and identifying its own mistakes. The system automatically collects modification work data when the robot corrects its own errors, minimizing the need for continuous operator intervention. This self-learning mechanism accelerates data collection while ensuring high model accuracy through targeted learning from correction scenarios.
3Manufacturing precision
If the robot performs provisional operations with human intervention, then the quality of training data is improved, but the complexity of the control system increases
Solution Approach 1:
The patent introduces an intermediary mechanism where the system automatically manages the transition between provisional operation and modification work data collection. The control system includes dedicated sections for provisional operation information output, modification work data collection, and trained model building, which mediate between human input and robot learning. This structured intermediary approach ensures high training data quality while organizing system complexity into manageable functional modules.
Data Source
AI summary
A robot control device includes a modification work trained model building section. The modification work trained model building section builds a modification work trained model by training on modification work data when a user's modification operation is performed to intervene in a provisional operation of a robot arm to perform a series of operations. In the modification work data, input data is a state of the robot arm and its surroundings when the robot arm is operating and output data is data of the operation by a user for modifying the provisional operation or the modification operation of the robot arm by the user's operation for modifying the provisional operation.