Traversable Object Classification Using Free Space Probability Differences

The patent introduces map data as an intermediary that provides height information about objects. Instead of directly measuring height with expensive 3D radar, the system uses pre-existing map data (from satellite imagery, aerial photography, or other sources) that contains elevation information. This map data acts as a mediator that bridges the gap between the inexpensive 2D radar and the need for height information, allowing the system to distinguish traversable objects like curbs without using costly 3D sensing hardware.

The patent uses map data as a copy or representation of the real-world environment's height information. Instead of directly measuring heights with 3D radar, the system creates a digital copy of elevation data from map sources (satellite imagery, aerial photography) and uses this copied information for classification. This allows the system to obtain height information indirectly through data copying rather than direct physical measurement, significantly reducing hardware costs while maintaining classification accuracy.

The patent segments the classification process into two distinct stages: first detecting objects with radar, then classifying them as traversable or non-traversable using map data. Instead of treating all detections uniformly as obstacles, the system divides the problem into detection and classification phases. This segmentation allows the system to maintain safety by detecting all objects while avoiding false positives, because the classification step correctly identifies traversable objects like curbs that should not be treated as hazardous obstacles.