Efficient Evaluation of Imprecise Location-Dependent Queries (Technical Report)

In location-based services, it is common for a user to issue a query based on her current position. An example is the location-dependent range query, where the user’s location is at the center of the query range. Very often, the query issuer’s location is imprecise due to measurement error, sampling error, or message delay. She may also want to protect her privacy by providing a less precise location. In this paper, we study the efficiency of queries that return probabilistic guarantees for location data with uncertainty. We classify this query into two types, based on whether the data (1) has no uncertainty (e.g., shops and restaurants), or (2) has a controlled degree of uncertainty (e.g., moving vehicles). Based on this classification, we develop three methods to improve the computational and I/O performance. The first method expands the query range based on the query issuer’s uncertainty. The second method exchanges the roles of query and data. The third method exploits the fact that users may only be interested in answers with probabilities higher than some threshold. Experimental simulation over a realistic dataset reveals that our approaches improve the query performance significantly.