Aggregated Multi-Objective Optimization via Metric Comparability Grouping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optimization methods, such as single-objective and multi-objective optimizations, face challenges in handling incomparable metrics like time and cost, leading to inefficient solution evaluation and resource utilization.
Innovation Solution
A generating apparatus that groups metrics based on comparability judgments and evaluation values, using integer programming to determine compensable metric values within each group, and generates objective functions to represent relationships between subjects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If multi-objective optimization is used to provide solutions for multiple objective terms, then the solution quality improves, but the time and resources required to evaluate all optimized solutions increase significantly
Solution Approach 1:
The patent segments the set of all objective terms into multiple groups, where each group contains comparable objective terms. Instead of evaluating all possible Pareto solutions across all objectives simultaneously, the method performs multi-objective optimization separately for each group, significantly reducing the total number of solutions that need to be evaluated while maintaining solution quality for comparable metrics.
2Loss of time
If single-objective optimization is used to provide a single optimized solution, then the evaluation time reduces, but the solution may not be optimized for multiple terms with incomparable relationships
Solution Approach 1:
The patent divides multiple objective terms into separate groups based on comparability. Each group can be optimized independently using single-objective or multi-objective optimization methods appropriate for that group's characteristics, allowing efficient evaluation time while maintaining solution quality for each specific group of comparable metrics.
Solution Approach 2:
The patent dynamically selects the optimization approach (single-objective or multi-objective) and grouping strategy based on the comparability relationships among objective terms. This adaptive approach allows the system to adjust its evaluation methodology to match the specific characteristics of the optimization problem, balancing solution quality and evaluation efficiency.
3Device complexity
If all objective terms are aggregated into a single objective function, then the optimization process simplifies, but terms with incomparable relationships (e.g., time and cost) cannot be properly optimized
Solution Approach 1:
The patent segments objective terms into multiple groups based on comparability, preventing the aggregation of incomparable terms into a single objective function. Each group is optimized separately with appropriate weighting schemes, maintaining optimization accuracy for incomparable metrics while keeping the overall process manageable through systematic organization.
Solution Approach 2:
The patent introduces comparability judgments and grouping structures as intermediaries between the objective terms and the optimization process. These intermediaries organize objective terms into comparable groups, allowing the optimization algorithm to process each group appropriately without attempting to directly aggregate incomparable terms, thus preserving optimization accuracy.
Data Source
AI summary
Groups of metrics to form objective functions may be generated by obtaining a judgment that represents whether a first subject is comparable to a second subject, obtaining a plurality of evaluation values for each of the first subject and the second subject, each evaluation value corresponding to a metric among a plurality of metrics, comparing, for each metric, a corresponding evaluation value of a first subject to a corresponding evaluation value of the second subject, and generating one or more groups based on a result of the comparison and the judgment, each group of the one or more groups including at least one metric of the plurality of metrics, wherein metrics in each group are determined to be comparable with respect to evaluating subjects.