Graph Neural Network Route Search Using ETA Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing route search technologies face challenges in accurately predicting the estimated time of arrival (ETA) due to inefficiencies in expressing road data for deep learning models, leading to suboptimal route planning and increased travel times.
Innovation Solution
The proposed solution involves expressing a road network as a graph neural network, converting road information into a graph structure where links are represented as nodes and connectivity is represented as edges, enabling the generation of an ETA prediction model that incorporates global, node, and edge attributes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If road data is expressed in traditional formats for deep learning models, then the model structure is simpler, but the ETA prediction accuracy deteriorates
Solution Approach 1:
The patent transforms road data from traditional formats into graph neural network structures by changing the representation parameters - converting road links into nodes, connectivity into edges, and incorporating multi-dimensional attributes (global route attributes, node-level link attributes, edge-level connectivity attributes). This parameter transformation enables the model to capture complex road network relationships, significantly improving ETA prediction accuracy despite increased structural complexity.
2Measurement precision
If deep learning-based ETA prediction is implemented, then prediction accuracy improves, but the efficiency of road data processing deteriorates
Solution Approach 1:
The patent segments road data processing into three distinct graph attribute layers: global attributes (route-level features), node attributes (link-level features like past speed and passage time), and edge attributes (connectivity features). This segmentation allows parallel processing of different data types and enables efficient feature extraction at each level, improving overall processing efficiency while maintaining high prediction accuracy.
Solution Approach 2:
The patent performs preliminary conversion of road data into graph neural network structure before feeding it to the deep learning model. By pre-processing and organizing road data into the appropriate graph format with all necessary attributes during the data preparation phase, the actual model training and prediction processes become more efficient, as the model receives ready-to-use structured data without requiring complex on-the-fly transformations.
3Ease of manufacture
If traditional route search methods are used, then computational simplicity is maintained, but route optimization quality deteriorates
Solution Approach 1:
The patent introduces graph neural networks as an intermediary layer between traditional route search algorithms and the final route optimization. The GNN processes road data to extract meaningful features and patterns, transforming raw road information into enhanced input data for route search algorithms. This intermediary processing improves route optimization quality by providing richer contextual information while maintaining the conceptual simplicity of traditional route search methods.
Data Source
AI summary
An apparatus and a method for searching for a route using estimated time of arrival (ETA) prediction based on a graph neural network are provided. The apparatus includes a storage module configured to store digital map data. The apparatus includes a processor configured to perform a route search based on an ETA prediction model according to a route exploration request. The ETA prediction model is generated based on a graph neural network formed by converting road information into a graph.