Solution-Determination System for Combinatorial Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing solution-determination systems for combinatorial optimization problems face challenges in determining optimal solutions quickly and precisely, as searches are often concentrated around local solutions rather than global optima.
Innovation Solution
A solution-determination system comprising multiple optimization processing units that perform neighborhood searches, along with a selection unit that identifies high-performing units and alters the search areas of less efficient units based on their solutions, thereby dispersing search resources across the solution space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple optimization processing units perform neighborhood searches in parallel, then the solution search speed is improved, but the search results may become concentrated around local solutions rather than global optima
Solution Approach 1:
The solution space is divided into multiple search areas, with each optimization processing unit responsible for a specific segment. The selection unit dynamically assigns different search areas to different processing units, ensuring diverse exploration across the solution space rather than concentration around local solutions.
Solution Approach 2:
The search areas of optimization processing units are dynamically altered based on performance evaluation. The selection unit identifies units with better evaluation values and uses their search areas as references to alter other units' search areas, creating a dynamic adaptation mechanism that prevents stagnation at local optima.
2Reliability
If all optimization devices operate asynchronously and share the best solution, then cooperation is enhanced, but search resources become concentrated around the best local solution found so far
Solution Approach 1:
Different optimization processing units are assigned different search areas based on their individual performance and the global search state. Each unit maintains its own search area with specific local characteristics, allowing diverse exploration strategies while still cooperating through the selection unit's coordination.
Solution Approach 2:
The search areas of optimization processing units are altered as a parameter change mechanism. When units with better evaluation values are identified, their search area parameters are used to modify other units' search areas, enabling adaptive resource distribution that prevents concentration around single local solutions.
Data Source
AI summary
A solution-determination system 100 of the present invention includes: a plurality of optimization processing units 121 that perform solution-determination of a single optimization problem by neighborhood search; and a selection unit that identifies a first optimization processing unit which is at least one optimization processing unit, and second optimization processing units which are another plurality of optimization processing units, and selects some optimization processing units of the second optimization processing units as alteration-target optimization processing units whose search areas are to be altered, on the basis of solutions determined at the respective optimization processing units.