Database Reference
In-Depth Information
most of the information needed for formulating RFID location tracking
queries. Since the starting and ending time for each object is included
in the table, a query can be performed on an EPC in order to deter-
mine the history of locations for that objects. The precise time taken
for an object to move from one location to another can also be derived
because of the presence of the
tstart
and
tend
variables. This history
of locations can be sorted temporally in order to provide the history of
locations for any object. Similarly, missing objects at a location can be
obtained from the
OBJECTLOCATION
table by comparing the objects at
that location, with the set of objects at any time and location, where
they were previously known to be complete. The precise formulation of
these queries is provided in [68].
RFID Data Monitoring Queries:
It is also formulate containment
or observation queries for any particular snapshot by making use of the
CONTAINMENT
,
OBJECTLOCATION
and
OBSERVATION
queries. In addition
temporal joins can be performed between different objects by formulat-
ing queries which examine the overlap between their
tstart
and
tend
variables. For example, recursive containment can be easily queried with
this approach with the use of
CONTAINMENT
table, and temporal aggre-
gation can be performed on the number of items which passed through
a location at a given time, by making use of the
tstart
and
tend
at-
tributes of the
OBJECTLOCATION
table. Thus, the tables supported by
the DRER model are
expressive
and can support a wide range of SQL
queries [68].
We note that in order to transform the noisy RFID into the high
level semantic tables discussed above, which are consistent and non-
redundant, a number of rules need to be defined. These rules correspond
to
data filtering
,
location transformation
,and
data aggregation
.Wenote
that many relationship tables (such as
containment
tables) are not
ex-
plicitly
specified in the RFID data, and they need to be inferred and
aggregated, based on the observation patterns. A rule-based framework
is proposed in [68] in order to automate the transformation of primitive
events into semantically cleaner representations. For example, a
data
filtering
rule can be defined to scan the data within a sliding window in
order to determine if there are duplicates of the same event in multiple
readers. One of these can then be dropped. Similarly, when a new lo-
cation for an object is defined by a particular reader, the ending time
stamp for the last location is updated to the current time. An entry is
created in the
OBJECTLOCATION
table which a new starting time stamp,
which is the current time. The ending time-stamp for the new entry
is set to
UC (Until Changed)
.Thisisanexampleofa
location trans-
formation rule
.Anexampleofa
data aggregation
rule is one in which
