Projects 3D vehicle and object uncertainty into 2D planes to compute collision probability faster with less overestimation for real-time planning.
Weights object data by integration history so high-priority targets are fused within each cycle, improving association reliability under heavy sensor loads.
Crowd-based search regions stabilize multi-sensor fusion, improving vehicle object detection when closely spaced targets confuse low-resolution sensors.
When many objects are detected, evaluation-value comparison keeps the most relevant targets without full priority sorting, cutting processing time.
Attitude-based filter smoothing adapts to target orientation, suppressing detector and environmental errors in velocity detection.