Autonomous Vehicle Path Prediction Using High-Order Object Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing path prediction methods for autonomous driving systems fail to accurately account for high-order interactions between objects, which can influence path determination, particularly through indirect and complex social interactions.
Innovation Solution
A method using a high-order graph model to generate a weighted adjacency matrix and Mth-order adjacency matrix to capture indirect interrelations between objects, integrating object attribute information through a combination layer to predict paths, and output control signals for autonomous driving.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a learning model uses only direct interaction analysis between objects, then the model complexity remains manageable, but the path prediction accuracy deteriorates due to inability to capture indirect and high-order interactions
Solution Approach 1:
The patent segments the interaction analysis into multiple hierarchical levels: direct interactions (first-order), indirect interactions through one intermediary (second-order), and further indirect interactions (third-order and higher). This segmentation allows the model to systematically capture complex social interactions while maintaining manageable computational complexity through structured processing at each level.
Solution Approach 2:
The patent extends the interaction analysis from a single dimension (direct pairwise interactions) to multiple dimensions by incorporating high-order interactions through multiple hops. This dimensional extension enables the model to capture indirect effects and chain reactions in social interactions, significantly improving path prediction accuracy for autonomous driving scenarios.
2Measurement precision
If the model captures high-order indirect interactions between objects, then the path prediction accuracy improves, but the computational resources and processing time increase
Solution Approach 1:
The patent performs preliminary computation by pre-calculating and storing interaction patterns and weights at different hierarchical levels during the model training phase. This preliminary action enables the model to quickly retrieve and apply pre-computed interaction information during real-time path prediction, reducing computational processing time while maintaining high prediction accuracy.
Solution Approach 2:
The patent implements dynamic computation where the model adaptively determines the depth of interaction analysis required for each specific scenario. Rather than always computing all high-order interactions, the model dynamically adjusts the analysis depth based on scene complexity and prediction requirements, optimizing the balance between accuracy and processing time.
Data Source
AI summary
A method for controlling an autonomous driving of a vehicle is introduced. The method may comprise generating a weighted adjacency matrix from an object graph, based on a location relative to an anticipated collision point among objects. The method may further comprise, based on this matrix and a graph convolution layer of a high-order graph model, generating an Mth-order adjacency matrix, defining interrelations among objects across M number of hops, generating object attribute information from this matrix and feature information defining an attribute of an object, generating, based on a combination layer of the high-order graph model, cumulative feature information by integrating the object attribute information, embedding the cumulative feature information into the object graph, generating, based on the embedding and a predicted path model, an object's predicted path, and outputting, based on this predicted path, a signal for controlling the autonomous driving of the vehicle.