Robot Controller Machine Learning for Article Storage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robot systems face inefficiencies in determining optimal operation start conditions, leading to unnecessary motions and difficulties for inexperienced workers in setting appropriate start positions, especially when the initial position is incorrectly set.
Innovation Solution
A controller with a machine learning device that observes operation start condition data and conveyance state data to learn optimal robot operation start conditions, using a state observing section, judgment data acquiring section, and learning section to associate these variables and determine appropriate start positions for efficient article storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the operation start position is set by a skilled worker based on experience, then the robot can perform article storage operations, but the position may be incorrect leading to useless motions and increased cycle time
Solution Approach 1:
The system enables self-learning through the learning section that automatically determines optimal operation start positions by analyzing conveyance state data and judgment data, eliminating the need for manual setup by skilled workers and continuously improving efficiency through automated feedback
Solution Approach 2:
The judgment data acquiring section collects feedback information including cycle time, number of missed articles, torque, and vibration, which is then used by the learning section to refine operation start position decisions, creating a closed-loop system that continuously optimizes performance
2Ease of operation
If the operation start position is arbitrarily decided in advance, then the system can operate, but inexperienced workers cannot determine optimal positions and the robot may perform unnecessary motions
Solution Approach 1:
The learning section automatically determines optimal operation start positions by analyzing conveyance state data and judgment data, enabling the system to self-optimize without requiring inexperienced workers to have specialized knowledge or experience
Solution Approach 2:
The system replaces manual expert judgment with an automated machine learning approach that processes sensor data and judgment data to determine optimal operation parameters, substituting human expertise with an automated intelligent system
3Productivity
If the robot starts following articles at incorrect positions, then the system can operate, but the robot performs useless motions increasing energy consumption and robot load
Solution Approach 1:
The system performs preliminary analysis of conveyance state data and article positions before the robot starts following articles, allowing the learning section to pre-determine optimal operation start positions and prevent useless motions before they occur
Solution Approach 2:
The judgment data including torque and vibration measurements provides feedback on robot load and energy consumption, which the learning section uses to optimize operation start positions and minimize unnecessary robot motions and energy usage
Data Source
AI summary
A controller is provided with a machine learning device learning an operation start condition for storing motions for an article on the carrier device by means of the robot. The machine learning device observes operation start condition data showing the operation start condition and conveyance state data showing states of articles on the carrier device, as state variables indicating a current state of an environment. Further, the machine learning device acquires judgment data showing an appropriateness judgment result of the storing motion and learns the operation start condition in association with the conveyance state data, using the observed state variables and the acquired judgment data.