Coherent LiDAR Point Cloud Clustering for Single-Frame Object Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous driving systems face challenges in accurately segmenting point clouds into distinct objects due to limitations in determining velocities based on a single sensing frame, leading to incorrect identification and tracking of moving objects, especially when objects have different velocities or are closely positioned.
Innovation Solution
The implementation of Doppler-assisted segmentation using coherent lidars, which utilize phase information from reflected signals to determine radial velocities, enabling efficient clustering of point clouds into objects by fitting coordinates and velocities to rigid-body equations, even in a single frame, and verifying hypotheses with additional frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If velocity determination is based on a single sensing frame, then processing time is reduced, but measurement precision deteriorates leading to incorrect object identification
Solution Approach 1:
The patent segments the point cloud into multiple clusters based on velocity thresholds, where each cluster represents potential objects with similar velocity characteristics. This segmentation approach allows the system to process velocity data from a single sensing frame efficiently while maintaining measurement precision by comparing velocities across multiple identified clusters rather than relying on a single velocity value per object.
2Measurement precision
If multiple sensing frames are used to determine velocities, then measurement precision improves, but processing time increases
Solution Approach 1:
The patent performs preliminary velocity calculations within each sensing frame by comparing return points across multiple frames and identifying velocity patterns. By pre-processing the velocity data to identify clusters with similar velocity characteristics before final object identification, the system achieves high measurement precision while minimizing the time required for final processing, as the heavy computational work of velocity comparison is done in advance.
3Measurement precision
If velocity thresholds are used for object separation, then object identification accuracy improves, but device complexity increases
Solution Approach 1:
The patent changes the parameter space by introducing velocity as an additional dimension for object differentiation, alongside spatial coordinates. By establishing velocity thresholds that separate objects based on their motion characteristics, the system improves object identification accuracy without significantly increasing device complexity, as the threshold-based approach uses simple comparative operations rather than complex algorithms.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach allows for precise identification and tracking of objects by distinguishing objects with different velocities and spatially close objects, improving the accuracy of autonomous vehicle navigation and path determination.
Implementation Method 1
obtain, by a sensing system of an autonomous vehicle (AV), a plurality of return points, each return point comprising one or more velocity values and one or more coordinates of a reflecting region that reflects a signal emitted by the sensing system
Data Source
AI summary
Aspects and implementations of the present disclosure address shortcomings of the existing technology by enabling Doppler-assisted segmentation of points in a point cloud for efficient object identification and tracking in autonomous vehicle (AV) applications, by: obtaining, by a sensing system of the AV, a plurality of return points comprising one or more velocity values and one or more coordinates of a reflecting region that reflects a signal emitted by the sensing system, the one or more velocity values and the one or more coordinates obtained for the same instance of time, identifying that the set of the return points is associated with an object in an environment, and causing a driving path of the AV to be determined in view of the object.