Vehicle Occupancy Mapping With Expectation Areas for Ghost Target Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing occupancy maps in ADAS and autonomous driving systems struggle with accurately distinguishing between stationary and moving objects, leading to false identifications of obstacles as ghost targets, which impair the representation of drivable space due to measurement inaccuracies and computation errors.
Innovation Solution
A method involving the creation of expectation areas around detected objects, repositioning these areas based on subsequent sensor data, and removing entries if no object is detected within the area for a predetermined number of scans, using log odds ratios and stop criteria to clean up the occupancy map.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If object detections are accumulated in the occupancy map without verification, then the occupancy map is updated continuously, but ghost targets and traces appear due to slow-moving or temporarily stationary objects being misclassified
Solution Approach 1:
The system performs preliminary actions by defining expectation areas around detected objects before final occupancy map updates occur. These expectation areas serve as temporary holding zones where object detections are monitored across multiple measurement cycles before being committed to the occupancy map, preventing premature accumulation of false obstacles
Solution Approach 2:
The system implements feedback mechanisms by continuously monitoring expectation areas across successive measurement cycles. Object detections within expectation areas are tracked and verified over time, with the system feedback-looping to confirm whether detected objects remain stationary or begin moving, thereby distinguishing true obstacles from temporary detections
2Measurement precision
If the expectation area is continuously repositioned to follow detected objects, then tracking accuracy is improved, but computational effort increases and may cause false obstacles for slowly moving objects
Solution Approach 1:
The system applies partial action by repositioning expectation areas only when necessary - specifically when new object detections occur within existing expectation areas. Rather than continuously tracking all objects, the system selectively updates expectation areas based on actual detection events, reducing unnecessary computational operations while maintaining tracking accuracy for relevant objects
Solution Approach 2:
The system changes parameters by adjusting the position and boundaries of expectation areas dynamically based on detected object locations. The expectation area parameters (position, size, shape) are modified to follow detected objects, optimizing the search region for subsequent detections while adapting to object movement patterns
3Reliability
If all detected objects are added to the occupancy map, then the map reflects all potential obstacles, but moving objects create ghost targets that impair drivable space representation
Solution Approach 1:
The system extracts and separates the verification process from the main occupancy map update flow. By defining distinct expectation areas that temporarily hold object detections for verification, the system extracts false obstacles from the final occupancy map before they can contaminate the drivable space representation, while true obstacles remain in the map after verification
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer implemented method for determining an occupancy map in the vicinity of a vehicle comprises the following steps: successively acquiring sensor data of a sensor system, determining object detections based on the sensor data, overlaying the object detections in a spatial representation of the vicinity of the vehicle, defining, for an object detection of a first data acquisition process, an expectation area extending around the object detection, adjusting, if an object detection of a second data acquisition process is present within the expectation area, the position of the expectation area based on a difference between the position of the object detection of the first data acquisition process and the position of the object detection of the second data acquisition process, and removing an object detection of the expectation area from the occupancy map if no object detection can be determined in the expectation area for a predetermined number of successive data acquisition processes.