Occluded Object Tracking Using Estimated Locations in Driving Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in tracking objects that become occluded, as they move out of the sensor's field of view or are obscured by other objects, leading to potential collisions due to the inability to maintain knowledge of their position.
Innovation Solution
A method is implemented to manage occluded regions by defining a map of the driving area, detecting objects using sensors, determining when an object enters an occlusion area, and creating an estimated object location, which is dropped when the occlusion area is cleared, with options to increase precision of observation for objects entering or leaving the area and adjusting a 'keep alive' duration based on the object's trajectory and time elapsed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the vehicle uses sensors to detect objects in the environment, then the vehicle can identify objects and obstacles, but objects may become occluded and lost from tracking when they move behind other objects or out of the field of view
Solution Approach 1:
The system performs preliminary actions by defining occlusion areas in advance based on the vehicle's position and sensor field of view. When an object enters an occlusion area, the system proactively creates an estimated object location and maintains tracking information before the object is completely lost, rather than waiting for detection failure to occur.
Solution Approach 2:
The system creates a copy of the object tracking information in the form of an estimated object location when the original detection is lost. This estimated location serves as a backup representation that maintains the object's track data, allowing the system to continue tracking the object even when direct sensor detection is unavailable due to occlusion.
2Loss of information
If the vehicle maintains tracking of occluded objects by creating estimated object locations, then the vehicle can prevent loss of object information, but the system complexity increases due to additional processing requirements
Solution Approach 1:
The system applies local quality by creating estimated object locations specifically within defined occlusion areas rather than applying complex tracking algorithms universally. The estimation process is localized to regions where occlusion is anticipated, based on the vehicle's geometry and sensor field of view, reducing overall system complexity while maintaining information preservation where needed.
Solution Approach 2:
The system performs preliminary action by pre-defining occlusion areas based on the vehicle's position and sensor characteristics before objects become occluded. This allows the system to prepare tracking estimates in advance for specific regions, rather than implementing complex real-time analysis across the entire environment, thereby reducing processing requirements.
3Measurement precision
If the vehicle increases precision of observation for objects entering or leaving occlusion areas, then the vehicle can improve object detection accuracy, but the energy consumption and computational load increase
Solution Approach 1:
The system applies local quality by concentrating increased precision of observation only at the boundaries of occlusion areas where objects are entering or leaving, rather than maintaining high precision observation uniformly across the entire field of view. This localized approach to enhanced detection reduces overall energy consumption and computational load while maintaining accuracy where transitions occur.
Solution Approach 2:
The system uses periodic action by implementing increased precision observation at specific moments when objects are detected entering or leaving occlusion areas, rather than continuously maintaining high precision observation. This event-triggered approach to enhanced detection reduces energy consumption and processing requirements compared to continuous high-precision monitoring.
Data Source
AI summary
Exemplary embodiments described herein include systems and method to remove tracks from a tracker when the tracks are within an occluded region, including defining a map of a driving area; defining an occlusion area within the map; detecting an object using sensor data from one or more sensors; determining that the object entered the occlusion area; creating an estimated object location for the object within the occlusion area; dropping the estimated object location for the object when the occlusion area is cleared.