Managing Uncertainty in Moving-Object and Sensor Databases

Speaker: Reynold CHENG

Abstract

In a moving-object database system, locations of objects are constantly reported to the database. These location values are subsequently used to answer user queries. Due to continuous changes in locations, as well as limited resources (e.g., network bandwidth and battery power), it is infeasible for the database to keep track of the actual location of every moving object. Queries that use the stale values provided by the database can produce incorrect answers. However, if the degree of uncertainty between the actual location value and the database value is limited, one can place more confidence in answers to the queries. More generally, query answers can be augmented with probabilistic guarantees of the validity of the answers. To answer these probabilistic queries, different solutions are required, depending on the moving pattern of an object, as well as the nature of the query. We first investigate the interesting issue of modeling ``uncertainty'' for a moving-object. Based on the uncertainty models, we will discuss how uncertain location data can be queried, where algorithms for range queries and nearest-neighbor queries will be presented.

We observe that uncertainty management techniques for a moving-object database can be generalized to a vast class of sensor-based applications, which typically involve the monitoring of continuously changing entities (e.g., temperature and pressure). I will talk about a taxonomy of uncertainty models for general sensor data. These classes of models describe uncertainty in different levels of precision. Next, we will present a classification scheme for probabilistic queries for uncertain sensor data, and briefly discuss how probabilistic queries are executed in each class. We also examine the important issue of measuring the quality of answers to probabilistic queries.