Learning Control Initialization Using Response Characteristic Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing learning control systems require a significant number of actual operations and man-hours for adjustment to determine appropriate correction amounts, which is inefficient and time-consuming.
Innovation Solution
A control device and system that includes a command value generation unit, a learning computation unit, and an initial value determination unit, which generates a second command value by compensating a first command value with correction data, updates correction data based on feedback, and determines an initial value for correction data using a response characteristic and estimation of feedback values, reducing the need for repeated actual operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If learning control is applied to determine appropriate correction amounts, then control performance is improved, but the number of actual operations and man-hours for adjustment increase significantly
Solution Approach 1:
The patent applies preliminary action by performing simulation learning computation before actual learning control to determine an initial value for correction data. This preliminary simulation phase allows the system to pre-calculate correction amounts based on response characteristics, thereby reducing the number of actual operations needed during subsequent learning control execution and decreasing adjustment time while maintaining control performance
Solution Approach 2:
The patent uses copying by creating a simulated environment that replicates the actual control system's behavior. The simulation learning computation unit generates a model of the control object and performs learning computations in this virtual copy, allowing correction data to be determined without repeatedly operating the actual control object, thus reducing time loss while preserving the effectiveness of learning control
2Measurement precision
If correction data is updated through actual learning control operations, then accuracy of control is improved, but the number of times the control object must be operated increases
Solution Approach 1:
The system performs preliminary simulation learning computation to determine an initial correction data value before actual learning control begins. This preliminary action provides a head start in achieving accurate control, reducing the number of actual operations needed to reach the desired accuracy level while maintaining the precision benefits of full learning control
Solution Approach 2:
The patent introduces simulation learning computation as an intermediary between system setup and actual learning control. This intermediary phase computes correction data in a virtual environment, allowing the system to achieve accurate control with fewer actual operations on the physical control object, thereby improving productivity without sacrificing measurement precision
Data Source
AI summary
A control device generates a second command value by compensating a first command value output at every control cycle according to a predetermined pattern with a correction amount output at every control cycle according to correction data, updates the correction data based on a deviation between the first command value and a feedback value from the control object, and determines an initial value of the correction data. The control device acquires a response characteristic indicating a relationship between an assigned command value and a feedback value shown in the control object in response to the command value, estimates a feedback value to be shown in the control object based on a value obtained by compensating the first command value with temporary correction data and the response characteristic, and updates the temporary correction data based on a deviation between the first command value and the estimated feedback value.