Autonomous Vehicle Object Validation Using Dual Coordinate Frames
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle LiDAR systems often incorrectly identify external objects, leading to potential safety issues in autonomous driving scenarios.
Innovation Solution
A vehicle control apparatus and method that utilizes a sensor to generate virtual boxes representing external objects, determining an interest virtual box based on vehicle state, object location, and box size, and validates these boxes using multiple coordinate systems and hysteresis to ensure accurate tracking and decision-making.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If LiDAR is used to identify external objects, then object detection capability is improved, but object identification accuracy deteriorates due to incorrect identification
Solution Approach 1:
The patent segments the object identification process into multiple validation stages: initial virtual box generation, coordinate system transformation, contour point distribution analysis, and hysteresis validation. Each stage processes specific aspects of the object data independently, allowing comprehensive verification while maintaining detection capability.
Solution Approach 2:
The patent implements feedback mechanisms through hysteresis validation that compares current object detections with previous frames. The system uses validation results to adjust tracking decisions, creating a closed-loop system that continuously improves identification accuracy based on temporal consistency feedback.
2Measurement precision
If multiple validation criteria are applied to virtual boxes, then object tracking accuracy is improved, but system complexity increases
Solution Approach 1:
The validation system is segmented into distinct modular components: virtual box generation module, coordinate transformation module, contour analysis module, and hysteresis validation module. Each module performs a specific validation function, making the complex system manageable and maintainable through clear separation of concerns.
Solution Approach 2:
The patent performs preliminary coordinate system transformations and contour point calculations before final validation decisions. By preparing data in advance through coordinate system conversions and distribution analyses, the system reduces the complexity of the final validation step while maintaining high tracking accuracy.
3Measurement precision
If coordinate system transformations are performed, then validation precision is improved, but processing time increases
Solution Approach 1:
The patent performs coordinate system transformations as preliminary steps before validation. By pre-calculating contour point distributions in both vehicle-centered and object-centered coordinate systems, the system prepares all necessary transformed data in advance, enabling fast validation decisions without repeated transformations.
Solution Approach 2:
The patent creates copied representations of the same spatial data in different coordinate systems. Instead of repeatedly transforming the original data, the system generates copy versions of contour points in both vehicle-centered and object-centered coordinate systems, allowing simultaneous validation in multiple reference frames without repeated processing.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An apparatus for controlling autonomous driving of a vehicle may comprise a sensor that obtains virtual boxes corresponding to a plurality of external objects, and a processor. The processor may identify an interest virtual box from these virtual boxes based on factors such as the vehicle's operating state, the external object's location, or the size of the corresponding virtual box. The processor may determine a first distribution of contour points, centered on the interest virtual box, in a first coordinate system and a second distribution of the contour points in a second coordinate system, centered on the contour points. The interest virtual box may be validated based on criteria such as its heading direction, location, and the distributions of contour points. Based on this validation, the processor may determine whether to output the interest virtual box, generates a signal, and controls autonomous driving based on the signal.