Point Cloud Obstacle Tracking in Dense Dynamic Traffic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In unmanned driving technologies, dense dynamic obstacles such as crowds and traffic flow lead to matching errors during obstacle matching, resulting in inaccurate obstacle tracking and affecting subsequent path planning.
Innovation Solution
An obstacle tracking method that determines an obstacle aggregation region in the first point cloud, classifies obstacles into aggregated and non-aggregated categories, and uses group and non-group matching rules to improve matching accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If obstacle matching is performed for every two consecutive frames of laser point clouds, then obstacle tracking is achieved, but matching errors occur when dynamic obstacles are dense, resulting in inaccurate tracking results
Solution Approach 1:
The patent segments obstacles into two categories: aggregated obstacles (in dense regions) and non-aggregated obstacles (in sparse regions). This segmentation allows different matching strategies to be applied to different obstacle types, improving overall tracking accuracy in dense environments while maintaining efficiency for sparse obstacles.
Solution Approach 2:
The patent applies different matching rules (group matching rule for aggregated obstacles, non-group matching rule for non-aggregated obstacles) based on the local density of obstacles. This local quality approach ensures that the matching process is optimized for each specific region's characteristics, resolving the contradiction between tracking reliability and matching precision.
2Measurement precision
If group matching rule with stronger constraint condition is used for aggregated obstacles, then matching accuracy improves, but computational complexity increases
Solution Approach 1:
By segmenting obstacles into aggregated and non-aggregated categories, the patent applies computationally intensive group matching rules only where necessary (in dense regions), while using simpler non-group matching rules for sparse regions. This reduces overall computational complexity while maintaining high accuracy where needed.
Solution Approach 2:
The patent implements local quality by applying different matching rule complexities to different spatial regions based on obstacle density. High-complexity group matching is applied locally to aggregated obstacles, while low-complexity non-group matching is applied to non-aggregated obstacles, optimizing the balance between accuracy and computational load.
Data Source
AI summary
An obstacle tracking method and apparatus, a storage medium and an unmanned driving device are provided. An obstacle aggregation region in a first point cloud may be determined according to position information of each obstacle in the first point cloud acquired by the unmanned driving device. Then, an aggregated obstacle and a non-aggregated obstacle in the second point cloud acquired by the unmanned driving device are determined according to the obstacle aggregation region. In addition, a matching result of each aggregated obstacle is respectively determined based on a group matching rule, and a matching result of each non-aggregated obstacle is respectively determined based on a non-group matching rule. Finally, an obstacle tracking result is determined according to the matching result of each obstacle in the second point cloud.