Vector Object Contours for Autonomous Vehicle Collision Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle navigation systems often rely on bounding boxes for object representation, which fail to provide accurate representations, compromising safe navigation.
Innovation Solution
Utilizing vector-based representations generated by determining vectors from a center of a cluster of points to a perimeter, which can be connected to form a bounding contour, providing a more accurate geometry of objects for collision avoidance and route planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If bounding boxes are used for object representation, then device complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The object representation is segmented into multiple vectors radiating from a central point to the object perimeter. Each vector captures a specific direction and distance to the object boundary, allowing the system to represent complex object geometries through multiple simple directional measurements rather than a single bounding box, thus improving precision while maintaining computational efficiency.
2Measurement precision
If vector-based representations are used instead of bounding boxes, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent transitions from two-dimensional bounding box representations to a polar coordinate system with radial vectors extending from a central point. This dimensional transformation allows the system to capture object geometry in multiple angular directions, providing richer geometric information while using a structured vector format that remains computationally manageable.
3Reliability
If more detailed object representations are used, then collision avoidance reliability is improved, but processing speed deteriorates
Solution Approach 1:
The system uses a limited number of strategically positioned vectors (e.g., 8 cardinal and intercardinal directions) to represent object geometry, which provides sufficient precision for collision avoidance without the computational overhead of complete perimeter mapping. This partial representation approach achieves adequate reliability while maintaining processing efficiency.
Data Source
AI summary
Techniques for vector-based representation of objects from sensor data of an autonomous vehicle are described herein. The vector-based representations are generated by identifying intersections of vectors originating at a center of an object and extending radially outward with a perimeter of the object based on sensor data. The vector-based representation may be generated using the intersection points to generate an object representation for use in planning and collision avoidance for an autonomous vehicle.