Get documents about "Introduction to Moving Objects Databases
Document Sample
Introduction to
Moving Objects Databases
Questions about Moving Objects
Points Questions Regions Questions
People When did Bush meet Arafat?
Forests, rivers, How fast is the Amazon rain
Animals Determine trajectories of rivers forest shrinking?
birds, whales, … What is the minimal extent of
Which distance do they river X during the year?
traverse, at which speed?
Glaciers Does the polar ice cap grow?
Sattelites, Which sattelites will get close Does it move?
spacecraft, to the route of this spacecraft
Storms Where is tornado heading?
planets within the next 4 hours?
When will it reach Florida?
Cars: taxi- Which taxi is closest to a
cabs, trucks passenger request position? People, troops, Movements of the celts in the
armies 2nd centure B.C.
Which routes are used
regularly by trucks? Hannibal traversing the Alps.
Show his trajectory.
Air planes Were any two planes close to
a collision? Traffic jams How has been the size of the
traffic jams changing during the
Did planes cross the air
last hour?
territory of state X?
Oil spills Which parts of the coast will be
Ships Are any ships heading
touched tomorrow?
towards shallow areas?
Cancer Can we find in a series of X-ray
Military All kinds of military analyses
images a growing cancer?
vehicles
How fast does it grow? How big
was it on June 1, 1995?
Moving Objects Databases
Moving objects database is a database
that can represent and query moving
objects
Moving object is a geometry that
changes over time continuosly
How to extend database technology to
support moving object databases?
Outline of this lecture
Database Management Systems
Spatial Databases
Temporal Databases
Moving Objects in Databases
Database Management Systems
Database management system (DBMS) is a piece of
software that manages a database
Database is a repository of interrelated data items that are
central to the business of an enterprise/institution
Physical level:
Data organization on storage media (files)
Logical level:
Data model and query/data manipulation language
E.g., relational data model and SQL
External views:
Application-specific view of data from logical level
Database Management Systems
Classical database management systems were
conceived for relatively simple business
applications:
E.g., in the relational data model, simple data types
(integers, floating-point numbers, short text strings)
We would like to widen the scope:
Data: Images, geographic maps, music, videos, data
from scientific experiments, meteorological
measurements
Queries:
“Retrieve images containing shapes similar to a given one”
“Produce a map of rainfall over some terrain”
In order to formulate queries in a simple manner
and to process queries efficiently, we need to
extend data model and query language
Limitations of Classical Databases
No way to represent geometric shapes:
Representation of a region as a collection of
coordinates in numeric attributes is difficult to
handle
No way to represent development of entities
over time:
Past state of the world is not kept
No way to represent objects moving right
now:
Continuous updates are not feasible
Outline of this lecture
Database Management Systems
Spatial Databases
Temporal Databases
Moving Objects in Databases
Spatial Databases
Spatial database (DBMS) allows representing and
querying geometries in a natural way
Entities to be stored in a spatial database:
Cities, rivers, road networks, landmarks, boundaries of
countries, hospitals, subway stations, forests, corn fields
Fundamental abstractions:
Single objects: point, line, region
Spatially related collections of objects: partition, network
Extending Data Model and
Query Language: ADTs (ROSE Algebra)
Spatial data type (SDT) is a structure (e.g., region)
Algebra is a collection of abstract data types, e.g., STDs, with related
operations
Algebra must be closed under its operations!
The ROSE algebra:
STDs: points, region, line
Other data types: real, bool, …
Operations:
Contour: region line
Length: line real
Inside: points x region bool
Adjacent: region x region bool
Intersection: line x line points
Minus: region x region region
Extending Data Model and
Query Language: ATDs (ROSE Algebra)
We embed SDTs into a DBMS data model (e.g., into a relational model)
in the role of attribute types
Sample relations:
Cities (name: string, population: int, location: points)
Rivers (name: string, route: line)
Highways (name: string, route: line)
States (name: string, area: region)
Sample queries:
“What is the total population of cities in France?”
SELECT SUM(c.population)
FROM cities AS c, states AS s
WHERE c.location inside s.area AND s.name = ‘FRANCE’
“Return the part of the river Rhine that is within Germany”
SELECT intersection (r.route, s.area)
FROM rivers AS r, states AS s
WHERE r.name = ‘RHINE’ AND s.name = ‘GERMANY’
Extending Data Model and
Query Language: ATDs (ROSE Algebra)
Implementation strategy:
Data structures for SDTs
Algorithms for SDTs operations
Spatial index structures
Extension of the optimizer:
Cost functions for all methods
Statistics about the distributions of objects in space,
needed for selectivity estimation
Translation rules
User interface extensions to handle presentation of spatial data and
input of spatial values for querying
Some of these have become available in commercial
systems:
Good extensibility by attribute data types and operations:
Oracle, IBM DB2, Informix, MySQL
Outline of this lecture
Database Management Systems
Spatial Databases
Temporal Databases
Moving Objects in Databases
Temporal Databases
Standard DBMS describe the current state of the world
Name Department
Lisa Toys
A change in the world leads to update in the database
Previous state is lost!
Name Department
Lisa Books
Temporal Databases
Application can manage time itself:
Explicit attribute of type date or time
Sample relation:
Disadvantages of this approach:
Complex query formulation
Inefficient query processing
Temporal Databases
Temporal projection:
What is Bob‟s current salary?
select Salary
from Employee
where Name = „Bob‟ and
Start <= CURRENT_DATE and
Stop >= CURRENT_DATE
Temporal Databases
Sample relation:
Compute the salary history:
What is Bob‟s salary history?
Need to coalesce consecutive, value-equivalent tuples
Temporal Databases
Salary history in standard SQL
Temporal Databases
Computing salary history is easy now:
select Name, Salary, Start, Stop
from Employee1
where Name = „Bob‟
Temporal Databases
How to compute the history of salary and title (original
relation)?
Need temporal join
Temporal Databases
Need to analyze how each tuple of Employee1
overlaps each tuple of Employee2
Temporal Databases
Temporal join in standard SQL:
The Time Domain
Time is a one-dimensional space extending
from the past to the future
Different models of time:
Bounded vs. infinite
Discrete vs. continuous
Discrete and continuous time:
With discrete time, atomic time intervals
(chronons) are isomorfic to natural numbers
With continuous time, points in time are
isomorfic to real numbers
Time Dimension
Two kinds of time:
Valid time is the time in the real world when an event
occurs or a fact is valid
Transaction time is the time when a change is recorded in
the database or a time interval during which a particular
state of the database exists
Temporal database (DBMS) is a database (DBMS)
that supports valid time and/or transaction time
Extending Data Models
General approach:
Database facts are associated with timestamps
Alternatives:
Data model extended:
Relational vs. object-oriented
Granularity of facts:
Tuples/objects vs. attributes
Kind of timestamp:
a single chronon (instant)
a single time period (period)
a temporal element (periods)
Time dimension:
Valid time
Transaction time
Both dimensions
Extending Model: Valid-Time Relations
1. On January 1, Lisa started Name Department Time
to work in the toys department and is
expected to work there until January 15 Lisa Toys [1;15]
Name Department Time
2. On January 8, Lisa moved to
Lisa Toys [1;7]
the books department and is
expected to work there until January 15 Lisa Books [8;15]
3. On January 14, Lisa moved back Name Department Time
to the toys department and is
expected to work there until January 20 Lisa Toys [1;7]
Lisa Books [8;13]
Lisa Toys [14;20]
Extending Model:
Transaction-Time Relations
Name Department Time
1. On January 6, it was recorded
that Lisa worked in the toys department Lisa Toys [6; uc]
Name Department Time
2. On January 10, it was recorded that Lisa Toys [6;9]
Lisa had moved to the books department
Lisa Books [10;uc]
Name Department Time
2. On January 12, it was recorded that
Lisa had moved back Lisa Toys [6;9]
to the toys department
Lisa Books [10;11]
Lisa Toys [12;uc]
Extending Data Model:
Bitemporal Relations
1. On January 6, it was recorded that Lisa
Name Department Time
had started working in the toys
department on January 1 and was Lisa Toys [6; uc] x [1;15]
expected to work there until January 15
Name Department Time
2. On January 10, it was recorded that Lisa
Lisa Toys [6;9] x [1;15]
had moved to the books department on
January 8. She was still expected to Lisa Toys [10;uc] x [1;7]
work until January 15.
Lisa Books [10;uc] x [8;15]
Name Department Time
Lisa Toys [6;9] x [1;15]
2. On January 12, it was recorded that Lisa
would move back to the toys Lisa Toys [10;uc] x [1;7]
department on January 14 and Lisa Books [10;11] x [8;15]
would stay there until January 20
Lisa Books [12;uc] x [8;13]
Lisa Toys [12;uc] x [14;20]
Extending Data Model: BCDM model
Data model extended: relational
Granularity of facts: tuple
Model of time: bounded, discrete
Kind of timestamp: a single chronon
Time dimension: both dimensions
Name Department Time
Lisa Toys {(6,1), …, (6, 15), …, (9,1), …, (9, 15),
(10, 1), …, (10,7), …, (uc, 1), …, (uc, 7),
(12, 14), …, (12, 20), …, (uc, 14), … (uc, 20)}
Lisa Books {(10, 8), … , (10, 15), …, (11, 8), … , (11, 15),
(12, 8) , … (12, 15), …, (uc, 8), … , (uc, 13)}
Extending Query Language: TSQL2
TSQL2:
is a superset of SQL92
is based on BCDM model
supported by TimeDB (ETH Zurich)
Sample data definition command:
CREATE TABLE prescription (
name char(30),
drug char(30),
dosage char(30) )
AS VALID DAY AND TRANSACTION
Kinds of relations:
Snapshot
Valid-time
Transaction-time
Bitemporal
Extending Query Language: TSQL2
Sample queries:
“What has ever been prescribed any drugs?”
SELECT SNAPSHOT name
FROM prescription
“What drugs were prescribed to Lisa?”
SELECT drug
FROM prescription
WHERE name = ‘Lisa’
“What drugs have been prescribed together with aspirin?”
SELECT p1.name, p2.drug
FROM prescription AS p1, prescription AS p2
WHERE p1.drug = ‘aspirin’ AND p2.drug<>’aspirin’
AND p1.name = p2.name
“Which drugs were Lisa prescribed during 1999?”
SELECT p.drug
FROM prescription AS p
WHERE p.name = ‘Lisa’ AND VALID(p) OVERLAPS PERIOD ‘[1999]’ DAY
“What did the physician believe on September 10, 1998, was Lisa‟s prescription
history?”
SELECT drug
FROM prescription AS p
WHERE name = ‘Lisa’ AND TRANSACTION(p) OVERLAPS DATE ‘1998-09-10’
Outline of this lecture
Database Management Systems
Spatial Databases
Temporal Databases
Moving Objects in Databases
Moving Objects in Databases
Moving object is a geometry that
changes over time continuosly
Moving objects database is a database
that can represent and query moving
objects
Two perspectives on
moving objects databases:
Location management perspective
Spatio-temporal data perspective
Location Management Perspective
Location management database is a snapshot database that is used to:
maintain dynamically the locations of a set of currently moving objects
ask queries about the current or near-future positions of
the moving objects (monitoring)
Main task:
Keeping the positions of a set of objects in a database,
e.g., a set of taxi cars, up-to-date
Scenario:
Moving objects send their current positions to the database and
the database performs updates
Trade-off:
If updates performed very often, then
the error is small
the update load is very high
If updates performed less frequently, then
the update load is low
the error is large
Solution:
Additionaly, store speed and direction of moving objects
Use this information to predict near-future objects‟ positions
Spatio-Temporal Data Perspective
Spatio-temporal database is a spatial database that keeps history of
changes to spatial objects
Two basic questions:
What kinds of data are stored in spatio-temporal databases?
Point, line, region, network, partition
Which kinds of change may occur?
Discrete change: any kind of data
Continuous change: point and region
Moving object is a geometry that experience continuous changes
Two most important abstractions of moving objects:
Moving point is an abstraction of a physical object moving around in the
plane, for which only position is relevant
Moving region is an abstraction of an entity in the plane that changes its
position, extent, and shape
Extending Data Model and Query
Language:
Temporal Database with SDTs
Idea:
Use a temporal DBMS and extend it with spatial data types (SDTs).
Example: TSQL2/BCDM extended with the ROSE algebra
Sample data definition command:
CREATE TABLE real_estate (
owner char(30),
area region)
AS VALID DAY
Sample query:
“Show the properties adjacent to the property of Charles Smith as of March 17,
1977”
SELECT r2.area
FROM real_estate AS r1, real_estate AS r2
WHERE r1.owner = ‘Charles Smith’ AND
VALID(r1) OVERLAPS DATE ‘[1977-03-17]’ AND
r1.area ADJACENT r2.area
This approach supports discrete changes, but does not support moving objects,
i.e., continuous changes!
Extending Data Model and Query Language:
Spatio-Temporal Data Types
Idea:
Use spatio-temporal data types with suitable operations as attribute types in
a snapshot DBMS
Spatio-temporal data types: mpoint, mregion
Operations:
Intersection: mpoint x mregion mpoint
Distance: mpoint x mpoint mreal
Min: mreal real
Trajectory: mpoint line
Deftime: mpoint periods
At_time: mregion x instant region
At_time: mbool x instant bool
Length: line real
Adjacent: mregion x mregion mbool
This approach can manage both discrete and continuous
changes!
Extending Data Model and Query Language:
Spatio-Temporal Data Types
Sample relations:
real_estate (owner: char(30), area:mregion)
flight (id: string, from: string, to: string, route:
mpoint)
weather(id: string, kind: string, area: mregion)
Sample queries:
“Show the properties adjacent to the property of Charles Smith as of March
17, 1977”
SELECT at_time(r2.area, instant(‘March 17, 1977’))
FROM real_estate AS r1, real_estate AS r2
WHERE r1.owner = ‘Charles Smith’ AND
defined(adjacent(r1.area, r2.area),
instant(‘March 17, 1977’)) = ‘true’
”Retrieve any pairs of airplanes, which, during their flight, came closer to
each other than 500 meters”
SELECT f.id, g.id
FROM flight AS f, flight AS g
WHERE f.id <> g.id AND min(distance(f.route,g.route)) < 0.5
At what time was flight BA488 within the snowstorm with ID S16?”
SELECT deftime(intersection(f.route, w.area))
FROM flight AS f, weather AS w
WHERE f.id = ’BA488’ AND w.id = ’S16’
Summary
Moving object databases are databases that allow
efficiently representing and querying moving
objects
Moving object is a geometry that changes
continuously
Classical data models and query languages (and
their implementations) need to be extended to
support moving objects:
Spatial extension:
Spatial databases: SDTs with operations (ROSE algebra)
Temporal extension:
Temporal databases: BCDM data model and TSQL2
Spatio-temporal extensions:
Temporal databases with SDTs
Spatio-temporal data types with related operations
Location management database
Related docs
