Bayesian Control Parameter Tuning With Adaptive Search Bounds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The optimization of control strategies for technical systems, particularly nonlinear systems, is challenging due to high computational effort and the need for numerous measurement processes, as existing methods either require extensive interaction with the environment or rely on unsuitable model structures.
Innovation Solution
A Bayesian optimization method is employed to iteratively determine model parameters within defined value ranges, using a quality function modeled as a Gaussian process regression, which balances exploration and exploitation to minimize the number of measurements and optimize control strategies efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If reinforcement learning is used to find a suitable control strategy, then no knowledge about the environment is required, but the interaction time with the environment during the learning process is very high
Solution Approach 1:
The patent applies preliminary action by pre-defining a model structure that describes the system behavior before optimization begins. This model structure serves as a preliminary framework that guides the subsequent parameter optimization process, reducing the need for extensive environment interactions during learning.
Solution Approach 2:
The patent introduces a model structure as an intermediary between the reinforcement learning algorithm and the actual system. This model acts as a mediator that approximates system behavior, allowing the optimization to proceed with fewer direct environment interactions while maintaining adaptability.
2Productivity
If conventional model-based methods are used with a specified model structure, then parameter adaptation can be carried out easily and efficiently, but the selected model structure may be unsuitable and the optimization will not yield an optimal result
Solution Approach 1:
The patent applies dynamics by making the model structure adaptable during the optimization process. Instead of fixing the model structure beforehand, the system dynamically adjusts the model structure based on observed system behavior, allowing both efficient parameter adaptation and reliable optimization results.
Solution Approach 2:
The patent employs parameter changes by optimizing not only the model parameters but also the model structure itself. This involves changing structural parameters of the model to better fit the actual system behavior, thereby improving both adaptation efficiency and optimization reliability.
3Ease of manufacture
If Bayesian optimization is used to create a control model, then an efficient black-box optimizer is obtained, but due to the high number of modeling parameters, a large number of measurement processes are necessary and long training times are the rule
Solution Approach 1:
The patent applies segmentation by dividing the optimization process into distinct phases: model structure selection, initial parameter optimization, and fine-tuning. This segmentation allows the system to focus computational resources efficiently at each stage, reducing overall training time while maintaining ease of control model creation.
Solution Approach 2:
The patent uses preliminary action by pre-selecting candidate model structures and pre-processing measurement data before the main optimization process. This preliminary preparation reduces the computational burden during training, decreasing training time while preserving the efficiency of Bayesian optimization.
Data Source
AI summary
Methods for ascertaining a control strategy for a technical system using a Bayesian optimization method. The control strategy is created based on model parameters of a control model and is executable. The method includes providing a quality function whose shape corresponds to a regression function and that evaluates a quality of a controlling of the technical system based on model parameters; carrying out a Bayesian optimization method based on the quality function in order to iteratively ascertain a model parameter set having model parameters within a model parameter domain that indicates the permissible value ranges for the model parameters; and determining the model parameter domain for at least one of the model parameters as a function of an associated maximum a posteriori estimated value of the quality function.