Occluded Object Tracking With Uncertainty Updates for Autonomous Driving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Self-driving vehicles face challenges in maintaining object tracking when objects become occluded, which poses a risk to safe navigation and collision prevention.
Innovation Solution
A method and system for tracking objects through occluded regions by defining a map with occlusion areas, detecting objects using sensors, creating and maintaining object tracks within these areas, and updating predicted locations with uncertainty adjustments based on sensor data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the vehicle uses sensors to detect objects, then object detection capability is improved, but objects in occluded regions cannot be detected
Solution Approach 1:
The system performs preliminary actions by maintaining object tracks in occluded regions based on predicted locations and uncertainty calculations. When an object enters an occlusion area, the system continues to track it using probabilistic methods and updates the predicted location with uncertainty adjustments, allowing the object to be recovered when it exits the occlusion area.
2Reliability
If the vehicle maintains tracking of occluded objects, then collision prevention is improved, but system complexity increases
Solution Approach 1:
The system changes parameters by introducing uncertainty adjustments to predicted locations of occluded objects. The uncertainty is calculated based on the duration of occlusion and object characteristics, allowing the system to maintain simple tracking logic while improving reliability through probabilistic position estimation.
3Measurement precision
If the vehicle updates predicted location with uncertainty adjustments, then object tracking accuracy is improved, but processing time increases
Solution Approach 1:
The system applies partial action by updating uncertainty adjustments selectively based on occlusion duration and object type. Not all occluded objects require full uncertainty calculations, allowing the system to balance accuracy with processing efficiency by focusing computational resources on high-risk scenarios.
Data Source
AI summary
Exemplary embodiments include systems and methods to maintain tracking of an object that passes into an occluded area, including defining a map of a driving area; defining one or more occlusion areas within the map; detecting an object using sensor data from one or more sensors; creating an object track for the object detected using sensor data; determining that the object track entered one of the one or more occlusion areas; and maintaining the object track while the object track remains in the one or more occlusion areas.


