Segmented Predictive Control Models for Lower Memory Use
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing control devices require excessive memory for predicting future states of control targets and their environments, leading to potential failures in state prediction due to insufficient memory capacity.
Innovation Solution
A control device with a segmentation unit that breaks down the model constructed by the model construction unit, allowing for reduced memory usage in predicting future states by segmenting the model and calculating control policies based on the segmented model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the state of the control target and its surrounding environment is defined in more detail for accurate prediction, then prediction accuracy is improved, but memory requirement increases excessively
Solution Approach 1:
The patent divides the control target into multiple segments or regions, and models the state of each segment separately rather than modeling the entire control target as a single detailed state. This segmentation allows the system to maintain prediction accuracy for critical areas while reducing the overall memory requirement by not detailing every aspect of the entire system.
Solution Approach 2:
The patent extracts and focuses on only the essential state variables that are critical for prediction accuracy, while omitting or coarsely representing less important details. This extraction approach maintains the necessary prediction precision while significantly reducing the memory footprint by storing only the most relevant state information.
2Reliability
If a detailed model is used for future state prediction, then prediction reliability is improved, but device complexity increases
Solution Approach 1:
The patent segments the complex control target into multiple simpler sub-models, each representing a specific segment or aspect of the system. This segmentation reduces the complexity of individual models while maintaining overall prediction reliability through the collective behavior of multiple simplified models.
Solution Approach 2:
The patent applies different levels of modeling detail to different segments of the control target based on their importance and characteristics. Critical segments are modeled with higher detail and accuracy, while less critical segments use simpler models, thereby reducing overall model complexity while maintaining prediction reliability for essential system behaviors.
Data Source
AI summary
This control device 10 comprises a model construction unit 11 that constructs a model for simulating a control object 20, a problem subdivision unit 12 that subdivides the model constructed by the model construction unit 11, a control measure calculation unit 13 that predicts the future status of the control object 20 using the model subdivided by the problem subdivision unit 12 and that calculates a control measure for the control object 20 on the basis of the predicted future status, and an operation command generation unit 14 that generates operation commands to the control object 20 on the basis of the control measure calculated by the control measure calculation unit 13.