Location Leveling

1. Introduction

Location based services (lbs) is an emerging application of mobile data services (Kumar, 2006; Seydim, 2001a; Xu, 2006). Using lbs, users with location-aware mobile units (mus) can query about their surroundings anywhere and anytime through wireless service providers (wsps) and information content providers (icps). However, unlike traditional data information services built on wired networks, to successfully deploy such services requires more than just adding delivery functionality and scalability on the top of the supporting infrastructure. While this ubiquitous computing paradigm provides a great convenience for information access, its constraints have posed a variety of challenges in query processing of location-based services (Xu, 2006). Moreover as a growing number of traditional systems are adapted to become wireless, systems become more and more complex. Middleware solutions have been proven to be a great benefit in incorporation of delivery features with ubiquitous system platforms, system extensibility, targeted wireless devices, back-end legacy systems, data storage, and workflow needs. In this paper we explore a middleware solution to query processing in location-based services.

Processing a location dependent query (ldq) (Seydim, 2001, 2001a, 2002) is the key to the lbs. Individuals using mus, including laptops and pdas, may access data located at icps using database queries. The responses to the mobile queries depend on the location of the user. The ldq query processing is complicated by the possible heterogeneous location representations and mismatch of the location granularities between wsps and icps. This so called location granularity mismatch may negatively affect location dependent queries (ldq), and consequently lbss. Consider the ldq “find the closest restaurant” in an lbs. There are at least two data accesses needed to complete this query. In the first access, the location of the mu is obtained from the location service (lcs) at the wsp. Then the location value is bound to the original query and a location aware query (laq) “find the restaurants in <default or assigned range> centered at <current location provided by wsp>” is formed. The laq is then sent through wireless delivery channels to icps. The icp returns a response of “a list of restaurants in <the given range> centered at <current location represented by icps>”. A more detailed discussion of ldq processing can be found in the literature (Seydim, 2001, 2001a, 2002). Here two issues are raised beyond simply data delivery. First, the location representations of wsp and icp may be heterogeneous and mismatched. Examples are cellid provided by wsps and street address by icps. Direct translation to a common representation such as longitude/latitude pair has been proposed as a solution in previous work. However this method is not sufficient when legacy systems are incorporated. When direct translation is applied, the common representation will need to be indexed for efficient search. But it is obviously not an optimal solution. Also the inherent restriction of longitude/latitude pair does not allow topologic information, such as river in-between, in mountain, and the 3^rd floor in the mall. Secondly the granularities of location representations may not be identical. Cellid and address may not be equivalent in location granularity. If the term “closest” is interpreted to be within five miles, the query could be interpreted to be “within 5 miles” of a specific latitude/longitude or cell or zip code and the responses to the query would be different.

Thus a reconciliation solution must be made for this location mismatch for location dependent query processing. This paper presents a data model, location leveling (ll,) that solves the problem in a general pervasive mobile computing environment as a middleware conciliation solution.