Autonomous Vehicle Cluster Tracking for Motion-Distorted Sensor Scans
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately tracking objects due to occlusions and sensor measurement inaccuracies, which affect driving decisions and route planning.
Innovation Solution
A method and system for tracking objects using a surface matching algorithm to associate clusters from consecutive sensor spins, adjusting motion characteristics, and controlling the vehicle based on accurate motion data, including techniques like Iterative Closest Point (ICP) and Kalman filters to correct for motion distortion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If sensor data is collected multiple times per second to track objects, then tracking frequency and responsiveness are improved, but measurement accuracy deteriorates due to motion distortion and occlusions
Solution Approach 1:
The system performs preliminary actions by collecting sensor data multiple times per second at high frequency before processing. This allows the system to capture objects in different positions and orientations, providing multiple samples that can be processed to correct for motion distortion and occlusions, thereby maintaining measurement precision despite high tracking frequency
Solution Approach 2:
The system uses feedback by comparing sensor data from multiple consecutive scans and using surface matching algorithms to identify corresponding points. The system continuously refines object tracking by comparing current sensor data with previous data, correcting for motion distortion through iterative matching, and adjusting tracking based on the identified correspondences to maintain accuracy at high update rates
2Measurement precision
If surface matching algorithms are used to correct motion distortion, then measurement precision is improved, but computational complexity increases
Solution Approach 1:
The system applies segmentation by dividing the sensor data into clusters of points that correspond to different objects. By segmenting the data first, the surface matching algorithm only needs to process points within each cluster rather than the entire point cloud, significantly reducing computational complexity while maintaining measurement precision for each object
Solution Approach 2:
The system implements local quality by applying surface matching algorithms locally to each cluster of points rather than globally to all sensor data. This allows the computational effort to be focused on local regions containing objects of interest, improving measurement precision for those regions while reducing overall computational complexity by ignoring empty or irrelevant areas
3Stability of the object's composition
If clusters are associated between consecutive spins to track objects, then object tracking continuity is improved, but reliability decreases due to occlusions and segmentation errors
Solution Approach 1:
The system uses feedback by continuously comparing sensor data from consecutive spins and using surface matching algorithms to identify corresponding points between scans. The system maintains object tracks by finding correspondences between clusters in different scans, using the identified matches to correct for occlusions and segmentation errors, and continuously refining tracking based on accumulated evidence from multiple scans
Solution Approach 2:
The system performs preliminary actions by collecting and storing sensor data from multiple consecutive spins before making final tracking decisions. This allows the system to have multiple samples of object positions and to use surface matching to identify correspondences across scans, providing redundancy that improves reliability when occlusions or segmentation errors occur in any single scan
Data Source
AI summary
The technology relates to tracking objects in an environment around an autonomous vehicle. A computing system of the autonomous vehicle determines accurate motion characteristics of objects detected in its environment despite various sensor measurement limitations. By correcting motion distortion for fast moving objects and accounting for discrepancies in sensor data gathering, motion characteristics may be determined for the detected objects with enhanced accuracy. Multiple sets of correspondences are determined for clusters from multiple sensor spins, enabling better alignment using a surface matching algorithm even when clusters have fewer data points. Efficiency is also enhanced by selecting hypotheses based on confidence levels. These techniques provide for identifying the types of objects for which a yaw rate can be accurately determined. Object classification can also be improved by accumulating associated clusters corresponding to a detected object. In addition, under- or over-segmentation can be mitigated with such techniques.