Transcript to view the file

Managing Location Information
for Billions of Gizmos on the Move
– What’s in it for the Database Folks
Ralf Hartmut Güting
Fernuniversität Hagen, Germany
Panel: Managing Location Information ...
My background:
– spatial database systems
– some work on models and query languages for moving objects
DBMS technology should be extended as follows:
1. Describing moving objects (gizmos)
– Extending data models so that moving objects can be described (new data
types, in my view).
– Extending query languages so that all (well, many) kinds of questions about
moving objects and their relationships to static spatial objects can be
formulated.
ICDE 2001
Page 2
1. Describing Moving Objects
Static spatial object:
position: point
Moving object:
position: f: time  point
distance(mo, obj)
inside(mo, obj)
t
f: time  real
f: time  bool
y
Continuous functions must be handled in
DBMS models and languages.
ICDE 2001
x
Page 3
1. Describing Moving Objects
Specific challenges:
– Integrate proposals dealing with moving objects in the past with those
describing them at present/future.
– Integrate modeling and querying of networks with modeling of movement
(objects move in networks in many cases).
– Model aggregation of moving objects.
Given observations of (lots of) cars on highways, compute traffic jams.
– Integrate position uncertainty into modeling and querying. May come from
observations or from descriptions.
“On monday morning, I arrived in Heidelberg. I took a walk downtown. From
about 11am to 2pm I visited the castle. I then took a train to Munich ...”
ICDE 2001
Page 4
2. Location Dependent Queries
A person or device on the move issues a query, e.g.
“Finde the five Italian restaurants closest from here.”
In principle a normal spatial query (substitute current position for here), but ...
– Indexing might continuously adapt to current position. For example, restaurants
always ordered by distance.
– For a PDA with limited memory, a cache might be continuously updated to
contain the current environment information.
Query depending on moving location:
“Notify me as soon as we get within 5 kms of a gas station.”
Can also be viewed as a continuous query:
“Find gas stations within 5 kms from now on”
– stop query n
Result can also be dynamic due to movement of queried objects:
“How many police cars are in the city center?” from now on. “Notify me
whenever their number changes by more than 3.”
ICDE 2001
Page 5
3. More ...
Implementation issues:
– Indexing of current / past movement
– Algorithms for operations on spatio-temporal (= moving object) data types
– Mapping of abstract, continuous models into finitely representable, discrete
models
Problems of scale:
– Handle position updates for two million cars moving around in Germany,
reporting their position every ten seconds.
Distributed, localized management of information
– Allow uniform, integrated access to local, space-related information residing on
many heterogeneous servers.
ICDE 2001
Page 6