Vehicle Sensor Data Sharing for Occlusion Mitigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately recognizing their surrounding environment due to sensor occlusion caused by obstacles, which hinders the effective use of sensors like cameras and lidar.
Innovation Solution
A vehicle control method that shares low-level sensor information between nearby vehicles using high-definition map data to update sensor data maps and perform map matching, selecting target vehicles based on occlusion detection and confidence calculations to improve recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data is collected only by the host vehicle, then device complexity is reduced, but measurement precision deteriorates due to sensor occlusion
Solution Approach 1:
The patent uses neighboring vehicles as intermediaries to obtain sensor data about structures that are occluded from the host vehicle's perspective. Instead of directly solving the occlusion problem with complex sensors, the system mediates through other vehicles that have line-of-sight to the occluded structures, allowing the host vehicle to indirectly access this information through V2V communication.
Solution Approach 2:
The patent merges sensor data from multiple sources - the host vehicle's own sensors and neighboring vehicles' sensors - into a unified sensor data map. This combination allows the system to overcome individual vehicle limitations and achieve comprehensive environmental recognition that neither vehicle could achieve alone.
2Measurement precision
If sensor data is shared between multiple vehicles, then measurement precision improves, but loss of information increases due to data validation requirements
Solution Approach 1:
The patent implements a feedback mechanism where sensor data from neighboring vehicles is validated against the host vehicle's sensor data map. The system calculates confidence levels based on multiple factors (classification match, moving flag, data quality, measurement distance) and only accepts data that passes validation thresholds, creating a feedback loop that ensures data integrity while filtering out unreliable information.
Solution Approach 2:
The patent changes the parameter of data acceptance from binary (accept/reject) to probabilistic (confidence level). By calculating confidence scores based on multiple parameters and comparing them against thresholds, the system can make nuanced decisions about data validity, reducing information loss while maintaining accuracy standards.
3Productivity
If sensor beam sections are increased to cover more area, then productivity improves, but object-affected harmful factors increase due to occlusion
Solution Approach 1:
The patent adds a temporal and spatial dimension to sensor coverage by utilizing data from neighboring vehicles at different positions. Instead of relying solely on the host vehicle's static sensor field, the system incorporates sensor data from multiple vehicles moving through different spatial positions, effectively expanding the coverage dimension and reducing occlusion blind spots.
Data Source
AI summary
A vehicle control method according to an embodiment includes selecting a target vehicle to share low-level sensor information with based on the low-level sensor information obtained by a host vehicle and information on a structure included in high-definition map information, and receiving low-level sensor information of the structure from the target vehicle to update a sensor data map of the host vehicle, and performing map matching with the high-definition map information through feature points of a road extracted based on the updated sensor data map.