Vehicle-to-Infrastructure Perception Fusion Using Virtual Obstacles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data fusion methods for intelligent driving vehicles, which combine on-board and roadside perception data, are inefficient due to the need for multiple processing operations, leading to incomplete data collection and processing challenges.
Innovation Solution
A vehicle-to-infrastructure cooperation information processing method that generates virtual obstacle data for the target vehicle based on positioning data and integrates it with on-board perception data, allowing for simplified data fusion with roadside perception data to create comprehensive fusion data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional data fusion methods are used to combine on-board and roadside perception data, then comprehensive obstacle detection is achieved, but multiple complex processing operations are required leading to low efficiency
Solution Approach 1:
The patent applies preliminary action by pre-processing on-board perception data to generate virtual obstacle data representing the target vehicle before fusion. This virtual obstacle data is created in advance with standardized formatting, so that during actual data fusion operations, the system only needs to integrate this pre-prepared virtual data with roadside perception data, eliminating the need for complex real-time processing operations and significantly improving fusion efficiency
Solution Approach 2:
The patent introduces virtual obstacle data as an intermediary element between on-board perception systems and roadside perception systems. This virtual obstacle data acts as a mediator that standardizes and simplifies the interface between the two different perception systems, enabling straightforward data integration without requiring complex multi-step processing operations to handle the inherent complexities of fusing heterogeneous perception data
2Loss of information
If on-board perception system alone is used to collect road data, then data collection is simple, but information collection is incomplete
Solution Approach 1:
The patent merges on-board perception data with roadside perception data through data fusion. By combining the data from these two different sources, the system achieves complete road environment information coverage - the on-board system provides data from the vehicle's perspective while the roadside system provides additional contextual information, together eliminating information loss without requiring complex multi-system coordination
3Loss of time
If virtual obstacle data is added to on-board perception data, then data fusion is simplified and speeded up, but additional data processing step is introduced
Solution Approach 1:
The patent applies preliminary action by creating virtual obstacle data in advance as a representation of the target vehicle in the on-board perception data. This pre-generated virtual data is then simply added to the on-board perception dataset before fusion, transforming what would otherwise be a complex real-time processing task into a straightforward data integration operation, thereby significantly reducing data fusion time despite adding one preprocessing step
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
The application discloses a vehicle-to-infrastructure cooperation information processing method, apparatus, device and an autonomous vehicle, and relates to the technical fields of intelligent transportation, autonomous driving and vehicle-to-infrastructure cooperation. The method includes: acquiring first on-board perception data of a target vehicle, the first on-board perception data including data of an obstacle around the target vehicle sensed by the target vehicle; generating virtual obstacle data for representing the target vehicle according to positioning data of the target vehicle; generating second on-board perception data based on the virtual obstacle data and the first on-board perception data; and fusing the second on-board perception data with roadside perception data to obtain fusion data, the fusion data including obstacle data of all obstacles in both the second on-board perception data and the roadside perception data, and the obstacle data in the roadside perception data including obstacle data for representing the target vehicle. The embodiment of the present application is applied to a main vehicle discovery and full perception fusion framework, and can simplify operation process of data fusion.