Vehicle Motion Prediction for Blind-Area Traffic Scenarios
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion prediction systems fail to accurately predict the motion of traffic participants in blind areas surrounding a vehicle, which are not detected by sensors due to obstruction, leading to incomplete situational awareness.
Innovation Solution
A motion prediction device that detects traffic participants using surrounding data, identifies blind areas, presumes the presence of virtual traffic participants in these areas, and predicts their motion based on traffic rules, using a processor with detection, determination, presumption, and prediction units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion prediction is based only on detected traffic participants, then prediction accuracy for visible participants is maintained, but prediction completeness deteriorates due to blind areas
Solution Approach 1:
The patent creates virtual traffic participants as copies or representations of potential real traffic participants in blind areas. These virtual participants are generated based on map data and traffic rules, allowing the system to predict motions that might occur in undetected regions without requiring direct sensor detection, thus compensating for information loss in blind areas.
Solution Approach 2:
The patent performs preliminary generation of virtual traffic participants before actual motion prediction occurs. By proactively creating these virtual participants based on predetermined map data and traffic rules, the system prepares for potential situations in blind areas, enabling more complete prediction coverage without waiting for actual detection or occurrence of events.
2Loss of information
If virtual traffic participants are presumed in blind areas, then prediction completeness is improved, but system complexity increases
Solution Approach 1:
The patent applies virtual traffic participant generation selectively only in blind areas where map data is available and traffic rules apply, rather than uniformly across all regions. This localized approach ensures prediction completeness in critical undetected zones while avoiding unnecessary complexity in areas already covered by sensor detection.
Solution Approach 2:
The patent uses a unified virtual traffic participant model that can represent multiple types of potential traffic participants (pedestrians, vehicles, cyclists) using the same computational framework. This multi-functional approach allows the system to handle diverse prediction scenarios in blind areas without requiring separate complex models for each participant type, thereby managing system complexity.
Data Source
AI summary
A motion prediction device detecting, from surrounding data representing a situation in a predetermined range of surroundings of a vehicle, a traffic participant existing in the predetermined range, determining whether there is a blind area not represented in the surrounding data in the predetermined range, when it is determined that there is a blind area presuming that there is a virtual traffic participant in the blind area, and predicting motion of the traffic participant caused by the presence of the virtual traffic participant.