Object Importance Determination in Autonomous Vehicle Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for predicting driver attention in autonomous vehicles are limited, as they rely solely on driving videos or images and struggle to capture all important information simultaneously, especially in complex driving scenarios, and do not effectively account for distractions or multiple relevant objects.
Innovation Solution
A method for determining object importance in vehicle control systems that uses a combination of visual and goal models to identify and prioritize objects based on their type and relevance to the vehicle's goal, allowing for more accurate vehicle control decisions by considering both the visual dynamics of road users and the vehicle's path.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional systems use only driving videos or images to predict driver attention, then the system complexity is low, but the prediction accuracy is insufficient in complex driving scenarios
Solution Approach 1:
The patent combines multiple input modalities (driving videos, images, and map data) with multiple processing models (visual model and goal model) to create a comprehensive object importance determination system. This merging of diverse data sources and models resolves the contradiction by achieving high prediction accuracy through multi-faceted analysis while managing complexity through integrated processing.
Solution Approach 2:
The system is designed to handle multiple types of inputs (visual data and map data) and perform multiple functions (object detection, importance determination, vehicle control). This multi-functionality allows the system to achieve accurate predictions across various driving scenarios without requiring separate specialized systems for each function.
2Loss of information
If the system captures all important information simultaneously using human gaze behavior, then the completeness of information is improved, but human gaze is sequential and cannot capture all information at the same time
Solution Approach 1:
The patent transitions from sequential temporal processing (human gaze) to parallel spatial processing by simultaneously analyzing multiple objects and information dimensions. The system processes visual data and map data in parallel, evaluating multiple objects concurrently rather than sequentially, thus achieving complete information capture without sacrificing processing speed.
Solution Approach 2:
The system performs preliminary processing of map data and visual data separately before integrating them for final object importance determination. This preliminary action allows the system to prepare multiple information streams in advance, enabling simultaneous comprehensive analysis without the sequential bottlenecks of human gaze behavior.
3Measurement precision
If the system filters objects based on object type and vehicle goal, then the decision-making accuracy is improved, but the processing time increases due to multiple evaluation criteria
Solution Approach 1:
The patent segments the object importance determination process into distinct components: visual model processing, goal model processing, and integrated evaluation. Each component handles specific aspects of the analysis independently, allowing parallel processing of different object attributes and reducing overall processing time while maintaining comprehensive evaluation accuracy.
Data Source
AI summary
Determining object importance in vehicle control systems can include obtaining, for a vehicle in operation, an image of a dynamic scene, identifying an object type associated with one or more objects in the image, determining, based on the object type and a goal associated with the vehicle, an importance metric associated with the one or more objects, and controlling the vehicle based at least in part on the importance metric associated with the one or more objects.