I have no expertise in this area, but my assumption is that this could help for a broader sort of object/world permanence for robots - e.g. if something is no l

I have no expertise in this area, but my assumption is that this could help for a broader sort of object/world permanence for robots - e.g. if something is no longer visible to the robot's sensors (e.g. behind an obstacle, smoke, etc) then it could use a model based on this type of tech to maintain a short-term estimate of its surroundings even when operating blind.