Robot Learning Control for Flexible Wire Arrangement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automation of flexible wire-like workpiece arrangement using industrial robots is challenging due to varying overall shapes and orientations of the workpieces, making it difficult to pre-program a universal operation for arrangement.
Innovation Solution
A machine learning device that acquires state variables such as the position, force, and posture of the workpiece and robot, and performs machine learning of algorithms to arrange flexible wire-like workpieces to a predetermined state, even when the shapes and orientations differ.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If pre-programmed operations are used for robot arrangement, then operational simplicity is maintained, but adaptability to varying workpiece shapes deteriorates
Solution Approach 1:
The system performs self-learning by automatically acquiring state variables from sensors and training data to generate arrangement algorithms without requiring manual programming for each workpiece variation. The robot learns through iterative training cycles where state variables are collected, algorithms are generated, and performance is evaluated, enabling the system to adapt to different workpiece shapes autonomously.
Solution Approach 2:
The system changes operational parameters dynamically by adjusting arrangement algorithms based on learned state variables. Instead of fixed pre-programmed operations, the robot modifies gripper position, orientation, transport path, and placement parameters according to the specific characteristics of each workpiece, enabling adaptability while maintaining operational efficiency.
2Adaptability or versatility
If machine learning with training data is implemented, then adaptability to different workpiece shapes is improved, but computational complexity increases
Solution Approach 1:
The machine learning process is segmented into distinct phases: data acquisition phase where state variables are collected from sensors, algorithm generation phase where training data is processed to create arrangement algorithms, and execution phase where learned algorithms are applied. This segmentation allows computational complexity to be managed in discrete steps rather than simultaneously, reducing overall system complexity.
Solution Approach 2:
The system performs preliminary actions by collecting and organizing training data in advance through the data acquisition unit. State variables including workpiece geometry, material properties, and arrangement requirements are gathered beforehand and stored as training data, so that the actual arrangement operation can proceed efficiently using pre-processed information rather than real-time complex computations.
3Manufacturing precision
If comprehensive state variables are acquired for accurate learning, then arrangement precision is improved, but data acquisition burden increases
Solution Approach 1:
The state variable acquisition system uses multi-functional sensors that simultaneously capture multiple parameters including workpiece position, orientation, gripper force, and transport trajectory. This universal acquisition approach reduces the data acquisition burden by consolidating multiple measurement functions into integrated sensing units rather than requiring separate specialized sensors for each parameter.
Solution Approach 2:
The system implements feedback mechanisms where acquired state variables are continuously monitored and used to adjust subsequent measurements and operations. Sensors provide real-time feedback on workpiece position and gripper force, allowing the system to refine data acquisition focus on critical parameters that most impact arrangement precision, reducing unnecessary data collection while maintaining high precision.
Data Source
AI summary
A machine learning device and a machine learning method are provided. A controller, which serves as the machine learning device and the machine learning method, is provided to perform machine learning of algorithms for arranging a flexible wire-like workpiece to a predetermined state using an industrial robot. In an embodiment, the machine learning device includes: an acquisition unit that acquires, as state variables, a state of the workpiece before arrangement starts and a state of the workpiece during arrangement; and a learning unit that performs machine learning of an algorithm for arranging the workpiece based on the state variables acquired by the acquisition unit.