Unmanned System Control with Rule-Based Schemes and DDPG Tuning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning models can construct highly accurate models that learn human-undetectable characteristics, but it's challenging to predict their output for various inputs, leading to uncontrollable and unsafe operations in apparatus control systems.
Innovation Solution
A control device that combines rule-based operation selection with a learning model, using an actor neural network and critic neural network updated by deep deterministic policy gradient (DDPG), to calculate and optimize operation parameters based on state information and sensor data, ensuring safe and reliable operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If a deep learning model is used to determine apparatus operations, then manufacturing precision and productivity are improved, but reliability deteriorates due to uncontrollable and unpredictable operations
Solution Approach 1:
The control system is segmented into two distinct components: a rule base that ensures reliable and controllable operations based on predefined rules, and a deep learning model that optimizes operation parameters for high precision. This segmentation allows each component to fulfill its strength while mitigating the weaknesses of the other.
Solution Approach 2:
The invention merges the rule base and deep learning model into a unified control system where the rule base determines the operation scheme and the deep learning model calculates optimal parameters within that scheme. This combination enables the system to achieve both reliability from rule-based control and precision from learning-based optimization.
2Productivity
If a deep learning model is used to determine apparatus operations, then productivity is improved, but device complexity increases due to the need for both rule base and learning model
Solution Approach 1:
The control system is divided into functional modules: a rule base module for determining operation schemes, a deep learning model module for parameter optimization, and a coordination module for integrating both. This modular segmentation manages complexity by organizing components with clear interfaces and responsibilities.
Solution Approach 2:
The system dynamically switches between rule-based control and learning-based optimization depending on the operational context. The rule base provides stable foundational control while the deep learning model dynamically adjusts parameters to optimize performance, creating a flexible yet manageable system architecture.
3Reliability
If rule base is used to determine operation schemes, then reliability is improved, but adaptability deteriorates due to inability to handle complex or unpredictable situations
Solution Approach 1:
The deep learning model continuously learns from operational data and feedback, enabling the system to adapt to new situations and complex scenarios while maintaining the reliability framework provided by the rule base. This feedback mechanism allows the system to evolve and handle unpredictable situations without compromising safety.
Solution Approach 2:
The rule base establishes preliminary operational frameworks and safety constraints before complex operations are executed. This preliminary structuring ensures that even when the deep learning model handles adaptive decision-making, operations remain within safe and reliable boundaries defined in advance.
Data Source
AI summary
A control device determines an operation scheme of an apparatus on a rule base on the basis of state information indicating the state of the apparatus or the state of an environment in which the apparatus operates, calculates a parameter for determining the content of an operation in the determined operation scheme on the basis of the state information and a learning model constructed on the basis of the operation of the apparatus based on the state information and evaluation of the operation, and commands the apparatus to execute the operation.