Siamese Neural Network for Selective Routing Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current routing optimization methods face challenges in efficiently identifying optimal routing solutions due to the vast number of candidate routes, leading to high computational costs and resource wastage on unpromising candidates, particularly in NP-hard problems like the Capacitated Vehicle Routing Problem (CVRP).
Innovation Solution
A computer-implemented method using a Siamese neural network to screen and identify promising routing solution candidates, coupled with a routing optimizer that applies reinforcement learning-based improvement operators, to selectively optimize routes and reduce computational burden.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of routing solution candidates are optimized to find the optimal routing solution, then the quality of the routing solution improves, but the computational cost and time consumption increase significantly
Solution Approach 1:
The value prediction network performs preliminary evaluation of routing solution candidates before they are subjected to full optimization. By predicting the cost or quality metric in advance, the system identifies promising candidates that warrant further optimization effort, while filtering out unpromising ones early in the process.
Solution Approach 2:
The value prediction network acts as an intermediary between the routing solution candidate generation and the full optimization process. This intermediary layer provides guidance on which candidates are worth optimizing, enabling selective optimization rather than exhaustive optimization of all candidates.
2Reliability
If the routing optimizer is applied to all routing solution candidates, then the likelihood of finding the optimal solution increases, but the computational resources are wasted on unpromising candidates
Solution Approach 1:
Instead of applying full optimization to all routing solution candidates, the system applies partial optimization only to those candidates identified as promising by the value prediction network. This selective approach uses exactly the amount of optimization effort needed - not more, not less - thereby avoiding waste on unpromising candidates while still finding optimal solutions when they exist.
Solution Approach 2:
The system changes the parameter of optimization application from universal (all candidates) to selective (only promising candidates). The value prediction network evaluates candidates and changes which ones undergo full optimization, transforming the optimization process from exhaustive to targeted.
3Measurement precision
If more routing solution candidates are generated and optimized, then the chance of identifying the optimal route increases, but the complexity of the system increases
Solution Approach 1:
The system segments the routing optimization process into two distinct stages: (1) value prediction using the neural network to evaluate candidates, and (2) full optimization only for promising candidates. This segmentation allows the system to handle large numbers of candidates efficiently by processing them in stages rather than all at once.
Solution Approach 2:
The system replaces the traditional mechanical approach of exhaustive optimization with an intelligent system based on value prediction. Instead of blindly optimizing all candidates, the neural network intelligently predicts which candidates are worth optimizing, substituting computational brute force with intelligent prediction.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for determining routing. An exemplary method comprises: inputting a plurality of to-be-optimized routing solution candidates to a Siamese neural network comprising a plurality of value prediction networks, each of the value prediction networks being trained to predict a cost associated with a to-be-optimized routing solution candidate; identifying one or more to-be-optimized routing solution candidates from the plurality of to-be-optimized routing solution candidates based on outputs of the Siamese neural network; inputting the one or more identified to-be-optimized routing solution candidates to a routing optimizer to obtain one or more optimized routing solution candidates; and determining an optimized routing solution with a lowest cost from the one or more optimized routing solution candidates.