Reward Function Generation for Reproducible Engine Control Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Creating control maps for engine control systems is a complex, individual-dependent process that relies heavily on expert experience, making it difficult for inexperienced engineers to generate appropriate control maps due to the lack of visualized evaluation criteria and the interdependence of manipulated variables, leading to variations in quality and increased time for adjustment.
Innovation Solution
A function generation program that uses manipulation data and measurement data to perform inverse reinforcement learning, generating a reward function with evaluation indices and coefficient distribution information, allowing for the creation of a standardized evaluation criterion that can be used to adjust control maps, thereby facilitating the creation of equivalent control maps by inexperienced engineers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If control maps are created based on expert experience and individual judgment, then the quality of control maps can be high, but the process becomes individual-dependent and difficult to reproduce for inexperienced engineers
Solution Approach 1:
The patent creates a standardized evaluation criterion that captures and replicates expert knowledge in a formalized structure. This criterion can be copied and applied by any engineer, transforming individual expert judgment into a reusable, standardized framework that ensures consistent control map quality across different engineers and organizations.
Solution Approach 2:
The patent transforms qualitative expert experience into quantitative parameters and evaluation indices. By changing the form of knowledge representation from subjective judgment to objective measurable parameters, the system enables inexperienced engineers to apply expert-level evaluation systematically through defined parameters and calculation formulas.
2Reliability
If multiple evaluation criteria are used to assess control maps, then the assessment becomes more comprehensive, but the complexity of the evaluation process increases
Solution Approach 1:
The patent divides the comprehensive evaluation into distinct segments: multiple evaluation indices (each assessing specific aspects like responsiveness, smoothness, fuel economy), calculation units for each index, and a weighted aggregation mechanism. This segmentation allows thorough evaluation while maintaining organizational clarity and reducing overall process complexity through structured modularity.
Solution Approach 2:
The patent merges multiple evaluation indices into a unified evaluation framework through weighted aggregation. By combining individual index results with appropriate weights into an overall evaluation score, the system achieves comprehensive assessment while simplifying the decision-making process through a single integrated metric that reflects all evaluation dimensions.
3Manufacturing precision
If control maps are adjusted manually by skilled persons, then the control quality can be optimized, but the time and man-hours required for adjustment increase significantly
Solution Approach 1:
The patent replaces manual mechanical adjustment processes with an automated information processing system. The calculation unit automatically computes evaluation indices and determines optimal control map parameters based on measured data and predefined criteria, substituting the mechanical manual adjustment process with automated computational processing that maintains optimization quality while dramatically improving efficiency.
Solution Approach 2:
The system enables self-service optimization where the evaluation criterion and calculation unit automatically assess control maps and guide adjustments without requiring continuous expert intervention. The standardized criterion serves as a self-contained guidance system that allows engineers to perform optimizations independently, reducing reliance on skilled persons and accelerating the adjustment process.
Data Source
AI summary
A non-transitory computer-readable recording medium stores a function generation program for causing a computer to execute a process, the process includes acquiring manipulation data generated based on manipulated variable distribution information that represents distribution of values of manipulated variables, and measurement data measured when a control object device is controlled based on the manipulation data, and by performing inverse reinforcement learning by using the manipulation data and the measurement data, generating a reward function that includes evaluation indices for the manipulated variable distribution information and coefficient distribution information that represents distribution of the values of coefficients of the evaluation indices.