From Position to Place

In From Position to Place, Jeffrey Hightower about the need for application to reason about “place”, not coordinates. Existing systems rely on manually defining places which, while useful, does not scale to ubiquitous deployment. Generically, place is a human-readable labeling of positions. Current approaches require manual definition of places. I must, by hand, delineate and label my neighborhood, property, rooms, furniture, and service areas of my devices. Manual definition does not scale. Instead, ubiquitous deployment requires automatically learning significant regions and semantically labeling them as places.

The challenge is to augment maps of physical features with the dynamic data to, over time, suggest geometric regions which are good candidates to label as places. Label directly represents the place�s demographic, environmental, historic, personal, or commercial significance and is the desired abstraction for emerging proactive applications.

The interface should be capable of answering both �What place labels are associated with my current coordinates?� and �What is the probability I am currently is in a place P?�