Safety Volume Lists for Multi-Hypothesis Object Detection

The patent segments the object detection process into multiple independent hypotheses, each associated with a different object recognition template. The system generates multiple detection hypotheses by comparing image information against multiple templates, then evaluates each hypothesis independently before combining results. This segmentation allows the system to consider multiple potential object identities without creating a monolithic complex system, as each hypothesis can be processed separately through the safety volume generation and motion planning pipeline.

The patent performs preliminary action by generating safety volumes for multiple detection hypotheses before final object identification is certain. Instead of waiting for a single definitive detection, the system proactively creates safety volumes for all plausible hypotheses and uses them in motion planning. This preliminary preparation of safety volumes for multiple scenarios improves reliability by ensuring robust motion planning regardless of which hypothesis proves correct, while managing complexity through standardized safety volume generation procedures.

The patent applies partial action by generating safety volumes for multiple detection hypotheses only when necessary - specifically when multiple object recognition templates match the image information or when unmatched regions are detected. The system does not always generate safety volumes for all possible hypotheses, but rather selectively does so based on the complexity of the detection scenario. This selective approach maintains motion planning robustness for challenging cases while avoiding unnecessary processing time for straightforward detections.

The system performs self-service through automated hypothesis evaluation and safety volume generation. Once multiple detection hypotheses are identified, the system automatically generates safety volumes for each hypothesis and integrates them into motion planning without requiring manual intervention or iterative refinement. This self-service capability reduces processing time by eliminating manual steps while maintaining the robustness benefits of considering multiple hypotheses.