Hierarchical Point Cloud Modeling for Articulated Vehicle Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous driving systems struggle to accurately model articulated objects, such as trucks with trailers, leading to incorrect estimation of their movement states due to representing them as single bounding boxes, which can result in overestimation of occupied area and deviation in trajectory predictions.
Innovation Solution
A method for dynamically generating a hierarchical representation of objects using point cloud frames, where a first group of frames generates an initial representation, and subsequent groups refine this representation based on subsets of points and associated confidences, allowing for accurate modeling of articulated objects by considering each part independently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If articulated objects are represented as single bounding boxes, then the representation process is simple, but the accuracy of movement state estimation deteriorates
Solution Approach 1:
The patent segments articulated objects into multiple independent parts (e.g., tractor and trailer for trucks, or front and rear vehicles for connected vehicles). Each part is represented by separate point cloud data and bounding boxes, allowing independent tracking of their positions and movements. This segmentation enables accurate estimation of each component's movement state while maintaining manageable complexity through modular processing.
2Measurement precision
If hierarchical representation with multiple parts is used, then the accuracy of object modeling is improved, but the computational complexity increases
Solution Approach 1:
The patent implements a dynamic hierarchical representation where the level of detail adapts based on observation confidence. When point cloud data provides high confidence in detecting object parts, the system creates detailed hierarchical representations with multiple bounding boxes. When confidence is low or data is insufficient, the system simplifies to coarser representations. This dynamic adjustment optimizes computational resources while maintaining modeling accuracy.
Solution Approach 2:
The patent applies partial action by selectively processing only those object parts that are confidently detected in point cloud frames. Rather than attempting to model every possible component of articulated objects, the system focuses computational effort on clearly identifiable parts, achieving sufficient accuracy without the full computational burden of exhaustive modeling.
3Speed
If static representation of objects is used, then the processing speed is fast, but the accuracy of movement state determination deteriorates
Solution Approach 1:
The patent employs periodic action by processing point cloud frames at regular time intervals to track object movements. Each frame provides a snapshot of object positions, and by comparing successive frames, the system determines movement states such as stationary, moving forward, reversing, or turning. This periodic updating maintains processing speed while capturing dynamic changes in object positions and orientations over time.
Data Source
AI summary
The present disclosure provides a method and apparatus for modeling an object. The method includes: obtaining a first group of point cloud frames associated with surroundings of a perception system; generating a hierarchical representation of an object in the surroundings based on the first group of point cloud frames; obtaining, from each frame of point cloud, a second group of point cloud frames associated with the surroundings, a group of points associated with the object and a first subset and a second subset of the group of points; determining, based on the first subset and the second subset, a first confidence and a second confidence; determining whether to change the hierarchical representation of the object based on the first confidence and the second confidence.