Spatial and Moving Objects Databases:

St t off th
State the Art
A t and
d Future
F t Research
R h Challenges
Ch ll

Markus Schneider
University off Florida
Department of Computer & Information
Science &Engineering
Abstract

Abstract
Spatial databases are full-fledged databases that, in addition, enable the storage,
retrieval, manipulation, querying, and analysis of geometries like points, lines, and
regions representing, for example, the geometries of cities, roads, and states
respectively. More complex examples are spatial partitions representing spatial
subdivisions like the counties in Florida and the election districts in Gainesville, and
spatial graphs representing spatial networks like transportation networks and
pipeline systems. Moving objects databases also deal with geometries but focus on
the change of their location and/or shape and/or extent over time. Examples
are moving points representing cell phone users, moving lines representing traffic
congestions, and moving regions representing hurricanes. The objective of this
tutorial is to highlight the state of the art of spatial and moving objects databases,
and indicate their future research challenges. The focus is on data models, querying,
data structures,
structures algorithms,
algorithms and system architectures.
architectures The tutorial will show that
spatial data types and spatiotemporal data types provide a fundamental abstraction
for modeling the geometric structure of objects in space, the temporally evolving
structure of objects in space and time, their relationships, properties, and operations.
Th
These data
d t ttypes d develop
l ththeir
i ffullll expressive
i power if th
they are iintegrated
t t d as
abstract data types into databases and their query languages.
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 2
Outline

Outline
1. Introduction
2 Spatial
2. S ti l D
Datat M
Modeling
d li
3. Spatiotemporal Data Modeling
4. Open Research Topics

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 3
Outline

Outline
1. Introduction
2 Spatial
2. S ti l D
Datat M
Modeling
d li
3. Spatiotemporal Data Modeling
4. Open Research Topics

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 4
Outline

Outline - Introduction
 Spatial and spatiotemporal data are ubiquitous
 Application
A li ti examples l
 Spatial and moving objects databases as interdisciplinary research fields
 Colloquial example queries
 Definition of spatial and spatiotemporal database systems

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 5
Introduction

Spatial and Spatiotemporal Data are Ubiquitous (I)

 80% of all available data have either an explicit or an implicit geographical
reference

 Explicit geographical references

 Actual geometries (spatial structures) of geographical objects
 Examples: city boundaries, lakes, railroads, landmarks, school districts

 Implicit geographical references

 Textual references to the names and descriptions of geographic objects
 Examples:
E l zip
i codes,
d street
t t names, city
it names, lland
d parcell numbers
b
 Geocoding: Google Map, MapQuest, Bing Maps, etc.

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 6
Introduction

Spatial and Spatiotemporal Data are Ubiquitous (II)

 Adding time lets spatial objects move
 Continuous
C ti evolution
l ti off geometries
t i over ti
time: movement,
t motion
ti
 Examples: hurricanes, traffic, whales, glaciers

 Conclusions
 Efficient and user friendly handling of large spatial data volumes is an
important
p task of database technologygy
 Understanding of spatial and spatiotemporal data is needed

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 7
Introduction

Applications for Spatial Databases (I)

Data management DNA helix Spatial analysis

foundation for
Geographical Information
Systems
y ((GIS))

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 8
Introduction

Applications for Spatial Databases (II)

The brain's complex network of The Bubble complex (the A street map
70 billion neurons and thousands Hodge object), a star cluster,
of kilometers of circuits in NGC 6946

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 9
Introduction

Applications for Moving Objects Databases (I)

Car navigation system

Traffic congestion
research

Migration routes of whales

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 10
Introduction

Applications for Moving Objects Databases (II)

Hurricane research Material

M t i l flow
fl in
i transportation
t t ti
pipelines (electricity, water, oil)

Movement of planets

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 11
Introduction

Spatial and Moving Objects Databases as

Interdisciplinary Research Fields
Spatial and Spatial
p and
Spatiotemporal Spatiotemporal
Database Data Warehousing
Systems Data Mining

Computational
C t ti l
Geographical Geometry
Information Spatial and
Systems (GIS) Moving Objects
Database Systems Artificial
Intelligence

Geosciences
Mathematics
Cognitive
Linguistics
S i
Science

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 12
Introduction

User Groups and Spatial Queries of Interest (I)

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 13
Introduction

User Groups and Spatial Queries of Interest (II)

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 14
Introduction

User Groups and Spatiotemporal Queries of Interest (I)

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 15
Introduction

User Groups and Spatiotemporal Queries of Interest (II)

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 16
Introduction

What is a Spatial Database System?

 A spatial database system is a database system.
 It offers spatial data types (SDTs) in its data model and query language.
 It supports spatial data types in its implementation, providing at least
spatial indexing and efficient algorithms for spatial joins.

From: R.H. Güting. An Introduction to Spatial Databases. VLDB Journal (Special Issue on Spatial Databases), 3(4):357-399,
1994.

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 17
Introduction

What is a Moving Objects Database System?

 A moving objects database system is a spatial database system.
 In addition, it offers spatiotemporal data types (STDTs) for moving objects
(including operations and predicates) in its data model and query
language.
 It supports spatiotemporal data types in its implementation, providing at
least spatiotemporal indexing and efficient algorithms for spatiotemporal
j
joins.

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 18
Introduction

Goals of this Tutorial

 Provide an overview and understanding of the essential concepts of
spatial and moving objects database systems
 What is special with spatial (geometric) data?
 What is special with spatiotemporal data?

 Emphasize the importance of spatial and spatiotemporal data types to

model and query (moving) geometries in databases

 Modeling and conceptual aspects are the focus of this tutorial but not
implementation aspects

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 19
Outline

Outline
1. Introduction
2 Spatial
2. S ti l D
Datat M
Modeling
d li
3. Spatiotemporal Data Modeling
4. Open Research Topics

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 20
Outline

Outline – Spatial Data Modeling

 How spatial data should not be represented
 Spatial
S ti l ddata
t ttypes
 Geometric set operations
 Metric operations
 Topological relationships
 Directional relationships
 Spatial querying

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 21
Spatial Data Modeling

How Spatial Data should NOT Be Represented (I)

 Example: Census block, and how we would perhaps like to store it in a
table

 Problem: Polyline is not a built-in data type

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 22
Spatial Data Modeling

How Spatial Data should NOT Be Represented (II)

 First solution ever proposed: Create a set of tables with shared attributes

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 23
Spatial Data Modeling

How Spatial Data should NOT Be Represented (III)

 Problems
 Polyline representation is scattered over several (!) tables
 How do we know that these four tables represent a polyline?
 The
Th relational
l ti l algebra
l b andd SQL do
d nott provide
id geometric
t i operations
ti
 SQL queries, if possible, are low-level queries
 Even if so
so, it would not be clear how to apply these geometric
operations to the tables
 Only chance
 Expensive joins needed to obtain full information about the polyline
 All tuples have to be loaded into main memory for further processing
 Tuples have to be transformed into a geometric data structure
 Geometric operations have to be applied to geometric data structure
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 24
Spatial Data Modeling

How Spatial Data should NOT Be Represented (IV)

 Conclusions
 This
Thi solution
l ti is
i highly
hi hl inefficient
i ffi i t
 Table representation does not allow the application of geometric
operations.
 Whole approach is unusable
 An object concept is needed
 Spatial and spatiotemporal data are/form complex application objects
and are not an accumulation of their constituent parts
 The relational data model and relational database systems cannot and
should not be used to represent complex application objects like spatial
and spatiotemporal data

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 25
Spatial Data Modeling

What are Spatial Data Types? (I)

 Spatial data types
 are special data types needed to model geometry and to suitably
represent geometric data in database systems
 Examples: point, line, region; partition (map), graph (network)
 provide a fundamental abstraction for modeling the geometric structure
of objects in space, their relationships, properties, and operations
 are an important part of the data model and the implementation of a
p
spatial DBMS
 The definition of SDTs
 is to a large degree responsible for a successful design of spatial data
models
 decisively affects the performance of spatial database systems
 exerts a great influence on the expressiveness of spatial query
languages
 should be independent from the data model used by a DBMS
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 26
Spatial Data Modeling

What are Spatial Data Types? (II)

 Conclusions
 An
A understanding
d t di off SDT SDTs iis a prerequisite
i it
 for an effective construction of important components of a spatial
database system like spatial index structures, optimizers for spatial
d t spatial
data, ti l query llanguages, storage
t management, t graphical
hi l user
interfaces
 for a cooperation with extensible DBMS providing spatial type
extension
t i packages
k like
lik spatial
ti l d
data
t blblades,
d cartridges
t id

 The definition and implementation of spatial data types is probably the

most fundamental issue in the development of spatial database
systems

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 27
Spatial Data Modeling

Which Spatial Data Need to Be Represented? (I)

 Two kinds of representations of spatial phenomena (vector/raster debate)
 objects
bj t in i space ((entity-oriented
tit i t d / feature-based
f t b d view)
i )
 vector data, boundary representation
 spatial database systems
 space itself
it lf (space-oriented
( i t d / position-based
iti b d view)
i )
 raster data, bitmap data, pixel data
 image database systems

 We consider
 modeling single
single, self-contained
self contained spatial objects
 modeling spatially related collections of spatial objects
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 28
Spatial Data Modeling

Which Spatial Data Need to Be Represented? (II)

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 29
Spatial Data Modeling

A Three-Level Model for Phenomena in Space

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 30
Spatial Data Modeling

Abstraction Levels for Data Type Design and

Implementation  Spatial/ST data types
 Abstract line, region, volume, surface, moving point,
moving region
Abstract Level  Definitions based on mathematics (e.g., point set
theory, point set topology, algebraic topology)
 Spatial/ST operations and predicates
 No consideration of implementation aspects

 Finite representations of spatial data types

 Discrete data structures for SDTs/STDTs
Discrete Level  Discrete line,, region,
g , volume,, surface , moving
g
point, moving region
 Algorithms for spatial/ST operations and predicates
on finite representations

 Geometric data structures

 In a programming language (C++, Java)
 Array structures, no pointer structures
Implementation Level  Geometric
G t i algorithms
l ith
 Methods from Computational Geometry

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 31
Spatial Data Modeling

Design Criteria for Modeling Spatial Data Types

General
Formal
Data model definition
definition
independence

Topological
Efficiency correctness
Design of
S ti l
Spatial
Data Types
Numerical
Extensibility
robustness

Closure Enabling finite

properties Geometric
representations
consistency

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 32
Spatial Data Modeling

Closure Properties
 General definition/structure of spatial objects

Spatial objects must be closed under set operations on the underlying

point sets

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 33
Spatial Data Modeling

Geometric Consistency
 Support of geometric consistency constraints for spatially related objects

SDT definition must offer facilities to enforce topological consistency

constraints

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 34
Spatial Data Modeling

Spatial Data Types – Intuitive Definition (I)

 Data type point: A (complex) point object consists of a finite number of
single points
points.
 Data type line: A (complex) line object consists of a finite number of
blocks (connected components). Each block consists of a finite number of
disjoint or meeting continuous curves.
curves
 Data type region: A (complex) region object consists of a finite number of
disjoint or 0-meet faces (sub-regions). Each face consists of an outer
cycle and a finite (zero
(zero, one
one, or more) number of disjoint or 00-meet
meet hole
cycles (holes).

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 35
Spatial Data Modeling

Spatial Data Types – Intuitive Definition (II)

 Advantages of these definitions
 Intersections
I t ti off two
t line
li objects
bj t can be
b represented
t d as a single
i l
complex point object
 No dependence of the object-oriented type constructors (set, list, array)
of the data model of the deployed database system
 No distribution of the single points of the result in several tuples of a
relational table
 Several disjoint rivers can be represented as a single line object
(waterways)
 Multiple components and holes in regions are needed since
 this guarantees the closure properties of spatial data types
 this is required by spatial applications (e.g., Italy with mainland,
offshore islands, and the Vatican)

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 36
Spatial Data Modeling

Why Do We Need a Formal Definition of SDTs?

 Better understanding of the complex semantics of spatial objects and
operations at the discrete level

 Formal definition of SDTs should be directly usable for a formal definition

of correspondingg spatial operations and predicates

 Clarity and consistency of the conceptual user view of SDTs

 A first step towards a standardization of spatial data types (achieved)

 Formal specification of SDTs for a possible implementation at the

implementation level

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 37
Spatial Data Modeling

Point Set Theory (I)

 Basic assumption: space is composed of infinitely many points (Euclidean
plane IR2) and contains a set of spatial objects
 Each spatial object Y can be regarded as the point set occupied by that
object (Y  IR2)
 Use
U off sett operations
ti , , − for
f constructing
t ti new spatial
ti l objects
bj t
 Possible anomalies

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 38
Spatial Data Modeling

Point Set Theory (II)

 Possible anomalies (continued)

 Ambiguities when defining topological relationships

 Let A and B two spatial objects, and let points be a function that
provides the point set of a spatial object
 A = B := points(A) = points(B)
A inside B := points(A)  points(B)
A intersects B := points(A)  points(B)  
 The definitions of = and inside are both covered by the definition of
intersects
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 39
Spatial Data Modeling

Point Set Topology (I)

 Same basic assumptions as point set theory but investigates topological
structures of a point set
boundary (Y), interior (Y o), closure (Y ), exterior (Y )

o Y
Y

Y

Y = Y o  Y Y o  Y =  Y   Y =  Y o  Y  = 
IR2 = Y o  Y  Y 

 Point set topology investigates properties that are independent of an

underlying distance or coordinate measure (metric) and that are
preserved under continuous topological transformations
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 40
Spatial Data Modeling

Point Set Topology (II)

 Regularization of point sets to avoid anomalies
 Spatial regions are defined as regular closed sets
o
 Y is a regular closed set if Y = Y
 Effect
Eff t off the
th interior
i t i function:
f ti elimination
li i ti off d
dangling
li points,
i t d dangling
li lilines
and boundary parts
 Effect of closure function: elimination of cuts and p
punctures by
y
appropriately adding points, adding missing boundary points
 Regularization function reg(Y) := Y o
 Geometric set operations are equated with regular set operations (A, B
regular closed sets)
A  B := reg
g(A  B) A  B := reg
g(A  B) g(A  B)
A ٓ B := reg

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 41
Spatial Data Modeling

Formal Definition of Spatial Data Types

 The spatial data type point is defined as

point = {P  IR2 | P is finite}

 The spatial data type line is defined as

 The spatial data type region is defined as

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 42
Spatial Data Modeling

Geometric Operations (I)

 Geometric union, geometric intersection, and geometric difference
(geometric set operations) of two spatial objects of the same spatial type
union, intersection, difference :      for all   {point, line, region}

 Formal definition the geometric set operations for  = region

 union(A, B) := A  B
: AB
 intersection(A, B) :=
 difference(A, B) := A ٓ B
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 43
Spatial Data Modeling

Geometric Operations (II)

 Geometric intersection of two spatial objects of different spatial data types
intersection:      for all ,
   {point, line, region} and   
 Computation of the common boundary of two spatial objects
commonBorder:     line for all ,   {line, region} and   

 Computation of the boundary of a spatial object

y: region
 boundary g  line
 boundary: line  point
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 44
Spatial Data Modeling

Geometric Operations (III)

 Computation of the common boundary points of two spatial objects
commonPoints:     point for all ,
   {line, region} and   

 Computation of the center of a spatial object

center:   point for all   {point, line, region}

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 45
Spatial Data Modeling

Geometric Operations (IV)

 Computation of the convex hull of a point object
convexHull: point  region

 Computation of the minimum bounding box (rectangle) of a spatial object,

that is, its smallest enclosing and axis-parallel rectangle
box:   region for all   {point, line, region}

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 46
Spatial Data Modeling

Geometric Operations (V)

 Zooming a spatial object by a factor
scale :   float   for all   {point, line, region}
 Rotation of a spatial object around a point by a given angle
rotate :   point  float   for all   {point, line, region}
 Moving a spatial object by a defined vector
translate :   point   for all   {point, line, region}
 Creation of a region object (a buffer) that contains all points in a certain
distance from a spatial input object (a negative distance lets the input
object shrink)
bufferZone :   float  region for all   {point, line, region}

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 47
Spatial Data Modeling

Geometric Operations (VI)

 Voronoi operation: For a given set S of points in the plane, the Voronoi
diagram associates with each point p from S the region consisting of
those points of the plane that are closer to p than to any other point in S

 The closest operation yields a point object whose point components are
nearest to a given single reference point
closest : point  point  point

closest point

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 48
Spatial Data Modeling

Metric Operations (I)

 Computation of the area of a region object
area: region  float

 Computation of the length of a region boundary

perimeter: region
g  float

 Computation of the length of a line object

length: line  float

 Computation of the largest Euclidean distance between any of two points

of a spatial object
 diameter: region  float
 diameter: line  float
 diameter: point  float
 In
I summary: diameter
di t :   float f allll   {point
fl t for i t, line
li , region
i }

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 49
Spatial Data Modeling

Metric Operations (II)

 Ideas for new unary distance operations
minIntraDist,
maxIntraDist,
minInterDist,
maxInterDist:   float for all   {point, line, region}

 Minimum/maximum Euclidean distance computation between two spatial

objects
dist, maxDist :     float for all ,   {point, line, region}

 Computation of the direction between two single points as the angle 0  

< 360 between them
direction : point  point  float

 Number of components (points, blocks, faces) of a spatial object

noOfComps:   integer

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 50
Spatial Data Modeling

Spatial Partitions and Their Operations (I)

 Spatial partition: subdivision of the plane into pairwise disjoint regions
where each region is associated with an attribute (label) having a simple
or even complex structure

 Partition constraints maintain topological

relationships between regions
 Interiors of different region objects
must be disjoint
 Different region objects are adjacent
(they meet) or they are disjoint

 Important
I t t example l off a spatial
ti l
connectivity structure

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 51
Spatial Data Modeling

Spatial Partitions and Their Operations (II)

 The overlay operation transparently lays two partitions of different themes
on top of each other and combines them into a new spatial partition
partition.

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 52
Spatial Data Modeling

Spatial Partitions and Their Operations (III)

 The superimposition operation superimposes the spatial objects of a
spatial partition on another spatial partition in order to cover and erase
parts of the other partition.

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 53
Spatial Data Modeling

Spatial Partitions and Their Operations (IV)

 The fusion operation merges the region objects of a spatial partition on the
basis of the equality of a non-spatial attribute
attribute. For each group of equal non-
spatial objects, a spatial object is created as the geometric union of the set of
spatial objects of each group.
 Example: Compute the spatial partition with the regions of the same land use

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 54
Spatial Data Modeling

Spatial Partitions and Their Operations (V)

 The cover operation yields a spatial partition with a single region object as
the geometric union of all region objects of a spatial partition
partition.

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 55
Spatial Data Modeling

Spatial Partitions and Their Operations (VI)

 The windowing operation retrieves those regions of a spatial partition
whose intersection with a given window (rectangle
(rectangle, circle
circle, polygon) is not
empty.

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 56
Spatial Data Modeling

Spatial Partitions and Their Operations (VII)

 The clipping operation selects those parts of a spatial partition which lie
inside a given window (rectangle
(rectangle, circle
circle, polygon)
polygon).

 Observation: The formal definition of all application-specific

application specific operations on
spatial partitions can be derived from three fundamental operations
 intersection
 relabel
 refine
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 57
Spatial Data Modeling

Spatial Networks and Their Operations (I)

 Spatial network: spatially embedded graph in the plane consisting of
pairwise disjoint,
disjoint continous lines where each line is associated with an
attribute (label) having a simple or even complex structure

 The shortest path (shortest route) operation

determines the shortest path between any two
network points with respect to
 network distance
 time
 slope
 any weight on which a distance metric can be
defined
 a combination of weights

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 58
Spatial Data Modeling

Spatial Networks and Their Operations (II)

 Windowing and clipping again, but now applied to spatial networks
(Euclidean range query)

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 59
Spatial Data Modeling

Spatial Networks and Their Operations (III)

 The operations spatial network union, spatial network intersection, and
spatial network difference compute the union
union, intersection
intersection, and difference
of two spatial networks of the same label type or different label types.
 Spatial network intersection: Given a bus route network and a taxi
network Which roads do they share?
network.
 Spatial network union: Given a bus route network and a metro network.
Which locations can a passenger reach?
 Spatial
S ti l network
t k difference
diff : Given
Gi ab
bus route
t network
t k and
d a taxi
t i
network. Which locations can taxis but not buses reach?

 The operation spatial network buffer considers a spatial network as a line

object and computes a buffer around it.
 Example: Given a bus route network. Compute the region that
represents its “zone
zone of attraction
attraction”.

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 60
Spatial Data Modeling

Spatial Networks and Their Operations (IV)

 The operation get graph extracts the graph structure of a spatial network.
 The
Th operation
ti network
t k distance
di t d t
determines
i th
the di
distance
t b
between
t ttwo
points in a network.
 Important for network range queries: Determine the spatial network that
h a di
has distance
t d from
f a network
t k point
i ts
 The operation length sums up the length of all edges of a spatial network
 What is the overall length of the bus network in Gainesville?
 The operation maxDist determines the maximum distance within a spatial
network
 What is the farthest distance in the bus network?
 The operation get transfer points determines the common points of two
different spatial networks where one can transfer from one network into
the other.
 Many more operations …
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 61
Spatial Data Modeling

Spatial Networks and Their Operations (V)

 Spatial network modeling in a database context
 Very few interesting models exist but they are incomplete
 Still a lot of effort needed
 Spatial network representation in a database context
 Poor: network data are spread over a large collection of tables
 Expensive joins needed to bring data together
 Network operations cannot be applied to these representations
 Spatial network querying in a database context
 Formulation of network queries is difficult
p
 No spatial network qqueries p
possible since network operations
p are missing
g
 High-level spatial network query language is missing
 Spatial network algorithms in a database context
 A number of efficient spatial network algorithms is available
 However: they depend on a particular spatial network database system
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 62
Spatial Data Modeling

Spatial Databases cannot Store Spatial Partitions

and Spatial Networks
 Both are highly complicated connectivity structures
 Both have complex operations involving large volumes of data
 Both are not first-class citizens in spatial databases

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 63
Spatial Data Modeling

Topological Relationships (I)

 Topological relationships characterize the relative position of two spatial
objects towards each other
 Examples:

meet disjoint inside

 Needed for spatial joins and spatial selections

SELECT cname, sname
FROM cities, states
WHERE loc inside territory

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 64
Spatial Data Modeling

Topological Relationships (II)

 Requirements of a model for topological relationships
 All topological
t l i l relationships
l ti hi mustt b be mutually
t ll exclusive
l i
 The set of topological relationships must be complete

 Proposed
P d models:
d l
 9-Intersection Matrix model (9IM)
 based on point set theory and point set topology
 takes into account the topological invariants of the intersections of
the boundary, interior, and exterior of one spatial object with the
corresponding components of another spatial object
 Region Connection Calculus (RCC)
 based on spatial logic, “pointless” approach
 axiomatic approach to formulating topological relationships and
reasoning about spatial data
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 65
Spatial Data Modeling

Topological Relationships (III)

 9-intersection matrix (A, B  {point, line, region})

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 66
Spatial Data Modeling

Topological Relationships (IV)

 Finding topological relationships: proof-by-constraint-and-drawing
 Phase 1
 Starting point: 29 = 512 possible matrices
d co
 Find constraint
s a rules
u es that
a must
us hold
o d for
o valid
a d topological
opo og ca relationships
e a o s ps
 Example: the intersection of the exteriors of two spatial objects is
always non-empty (eliminates 256 matrices immediately)
 Constraints are type-combination specific
 Apply constraint rules one after the other and remove matrices that do
not fulfill them
 Phase 2
 For the remaining matrices, draw prototypical spatial configurations in
the Euclidean plane to verify the existence of the corresponding
topological relationships

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 67
Spatial Data Modeling

Topological Relationships (V)

 Number of topological relationships for two simple spatial objects

 Number of topological relationships for two complex spatial objects

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 68
Spatial Data Modeling

Topological Relationships (VI)

 Example 1: Topological relationships between two complex point objects

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 69
Spatial Data Modeling

Topological Relationships (VII)

 Example 2: Topological relationships between two complex region objects

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 70
Spatial Data Modeling

Topological Relationships (VIII)

 Problem: too many topological relationships that the user cannot handle
 Solution:
S l ti clustering
l t i rules l ththatt putt similar
i il ttopological
l i l relationships
l ti hi ttogether
th
into a topological cluster predicate

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 71
Spatial Data Modeling

Directional Relationships (I)

 Directional relationships characterize the directional orientation of two
spatial objects towards each other
 absolute directional relationships (cardinal directions): north_of and
southwest_of with respect to a given reference or coordinate system
l ti directional
 relative di ti l relationships:
l ti hi i f t and
in_front d left
l ft with
ith respectt to
t an
observer position
p
 Needed for spatial jjoins and spatial
p selections
states(sname: varchar(50), territory: region)
SELECT s.sname
FROM states ss,
( SELECT territory
FROM states
y)t
WHERE sname = ‘Italy’)
WHERE s.territory northwest_of t.territory

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 72
Spatial Data Modeling

Directional Relationships (II)

 Objects Interaction Matrix (OIM) model
 Computational
C t ti l model
d l ffor cardinal
di l di
directions
ti
 Cognitive, subjective models can be built on top

 Two
T phases
h
 Tiling Phase
 Interpretation Phase

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 73
Spatial Data Modeling

Directional Relationships (III)

 Tiling Phase
 Determine
D t i Objects
Obj t Interaction
I t ti Grid
G id (OIG) through
th h partitioning
titi i lines
li
 Determine Objects Interaction Grid Space (OIGS)

A
OIG(A,B) =

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 74
Spatial Data Modeling

Directional Relationships (IV)

 Tiling Phase (continued)
 9 possible
ibl sizes
i off OIG
OIGs
 m × n-OIG for m, n ‫{ א‬1, 2, 3}

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 75
Spatial Data Modeling

Directional Relationships (V)

 Tiling Phase (continued)
 Map
M OIG into
i t an Objects
Obj t Interaction
I t ti Matrix
M t i (here:
(h 3 × 3-matrix)
3 ti )

 Objects Interaction Function

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 76
Spatial Data Modeling

Directional Relationships (VI)

 Tiling Phase (continued)
 Numbers
N b off possible
ibl m × n-matrices
ti

 Numbers of valid OIMs for two complex/simple region objects

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 77
Spatial Data Modeling

Directional Relationships (VII)

 Interpretation Phase
 Determine
D t i ththe llocation
ti off each
h partt off A and
d B in
i OIM(A, B)
 Example

A 1 1 0
 
 1 0 2 
0 2 2
 
B

 loc(A, OIM(A, B)) = {(1, 1), (1, 2), (2, 1)}

 loc(B, OIM(A, B)) = {(2,
{( , 3),
), (3,
( , 2),
), (3,
( , 3)}
)}

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 78
Spatial Data Modeling

Directional Relationships (VIII)

 Interpretation Phase (continued)
 Set
S t off basic
b i cardinal
di l di
directions
ti CD = {N, NW, W, SW, S, SE, E, NE, O}
 Determine the cardinal directions between the parts of A and B by an
interpretation function 

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 79
Spatial Data Modeling

Directional Relationships (IX)

 Interpretation Phase (continued)
 Interpretation
I t t ti table
t bl for
f the
th interpretation
i t ti 
t ti function
f

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 80
Spatial Data Modeling

Directional Relationships (X)

 Interpretation Phase (continued)
 Example
E l (continued
ti d)

A 1 1 0
  ((1, 1), (2, 3)) = NW
 1 0 2  ((1, 2), (3, 2)) = N
0 2 2
B   …

 Determine the cardinal directions dir(A, B) between A and B

dir(A, B) = (loc(A, OIM(A, B)), loc(B, OIM(A, B)))

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 81
Spatial Data Modeling

Directional Relationships (XI)

 Interpretation Phase (continued)
 Example
E l (continued
ti d)
dir(A, B) = (loc(A, OIM(A, B)), loc(B, OIM(A, B)))
A = ({(1, 1), (1, 2), (2, 1)},
{(2, 3), (3, 2), (3, 3)})
= {((1, 1), (2, 3)), ((1, 1), (3, 2)),
((1, 1), (3, 3)), …, ((2, 1), (3, 3))}
B = {NW, N, W}

dir(B, A) = (loc(B, OIM(A, B)), loc(A, OIM(A, B)))

1 1 0 = ({(2,
({(2 3),
3) (3,
(3 2)
2), (3
(3, 3)}
3)},
  = {(1, 1), (1, 2), (2, 1)})
 1 0 2 
0 2 2 = {((2, 3), (1, 1)), ((2, 3), (1, 2)),
  ((2,
((2 3)
3), (2 1)), …, ((3,
(2, 1)) ((3 3)
3), (2
(2, 1))}
= {SE, S, E}
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 82
Spatial Data Modeling

Directional Relationships (XII)

 Existential predicates

 dir(A, B) = {NW, N, W} means

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 83
Spatial Data Modeling

Directional Relationships (XIII)

 Similarly-oriented directional predicates

 Strict directional predicates

 Surround directional predicates

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 84
Spatial Data Modeling

Spatial Querying (I)

 Tables
 CREATE TABLE Country(name varchar(30),
varchar(30) pop int,
int shape region);
 CREATE TABLE River(name varchar(30), shape line);
 CREATE TABLE City(name varchar(30), pop int, shape point);
 Q1: Find the names of all countries that are neighbors of the USA.
SELECT C1.name AS neighbors
FROM Country C1, Country C2
WHERE C1.shape meets C2.shape AND C2.name = ‘USA’
 Q2: The St. Lawrence River can supply water to cities that are within 300
km Determine those cities and compute their distance from the river
km. river.
SELECT C.name, dist(C.shape, R.shape) AS distance
FROM City C, River R
WHERE C1 shape inside bufferZone(R.shape,
C1.shape bufferZone(R shape 300) AND
R.name = ‘St. Lawrence River’
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 85
Spatial Data Modeling

Spatial Querying (II)

 Q3: List the length of the rivers in each of the countries they pass through
together with the river name and the country name
name.
SELECT R.name, C.name,
length(intersection(R.shape, C.shape)) AS rLength
FROM River R, Country C
WHERE R.shape intersects C.shape

 Q4: List all countries, ordered by the number of adjacent countries.

SELECT C.name,, COUNT(C1.name)
( ) AS noOfNeighbors
g
FROM Country C, Country C1
WHERE C.shape meets C1.shape
GROUP BY C.name
ORDER BY noOfNeighbors DESC

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 86
Spatial Data Modeling

Spatial Querying (III)

 Q5: List the names of the pairs of all countries that are separated by
exactly one other country
country.
SELECT C1.name, C3.name
FROM Country C1, Country C2, Country C3
WHERE C1.shape meets C2.shape AND C2.shape meets C3.shape
AND C1.shape disjoint C3.shape

 Q6: List the names of all countries that are north of “Mexico”.
SELECT C.name
FROM Country C, Country C1
WHERE C.shape north_of C1.shape AND C1.name = ‘Mexico’
Effect of strictly_north_of,
strictly north of northern_of,
northern of strictly_northern_of?
strictly northern of?

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 87
Spatial Data Modeling

Spatial Querying (IV)

 Spatial data types
 are specified
ifi d as abstract
b t t data
d t types
t (ADT )
(ADTs)
 ADTs are used as attribute types in relational (or other) schemas
 Internals of ADTs are invisible to the user
 ADT implementations can be exchanged without the need to redefine
database schemas and reformulate queries

 Spatial operations and predicates

 are high-level methods
 are embedded into an extension of the SQL query language
 Algorithms for the operations and predicates are invisible to the user
 Algorithms for the operations and predicates can be exchanged without
the need to reformulate queries since the interfaces (signatures) stay
the same
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 88
Outline

Outline
1. Introduction
2 Spatial
2. S ti l D
Datat M
Modeling
d li
3. Spatiotemporal Data Modeling
4. Open Research Topics

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 89
Outline

Outline – Spatiotemporal Data Modeling

 What are Spatiotemporal Data Types?
 Uncertain
U t i MMoving
i Obj
Objects
t
 Predictive Moving Objects
 Spatiotemporal Partitions and Moving Objects in Spatial Networks
 Spatiotemporal Operations
 Spatiotemporal Predicates
 Spatiotemporal Querying

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 90
Spatiotemporal Data Modeling

What are Spatiotemporal Data Types? (I)

 Spatiotemporal data types are data types to represent moving objects.
 A moving object represents the continuous evolution of a spatial object
over time.
 Moving point
 Only time-dependent location is of interest
 Examples: eyes of tropical cyclones, cell phone users, terrorists,
whales
 Moving region
 Also the time-dependent shape and/or areal extent is of interest
 Examples: hurricanes, forest fires, oil spills, diseases, glaciers
 Moving line
 The time-dependent shape and/or linear extent is of interest
 Examples: traffic jams, front of an army; boundary of any moving region
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 91
Spatiotemporal Data Modeling

What are Spatiotemporal Data Types? (II)

 Formally: Let   {point, line, region}, and let time be the data type to
represent time instants
instants. Then a (historical) moving object m is a function
of a spatiotemporal data type () = time  .
 hmpoint = (point) = time  point
i = ((region
 hmregion
h i ) = time
ti  region
i
 hmline = (line) = time  line

simple moving point simple moving region complex moving point complex moving region

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 92
Spatiotemporal Data Modeling

Uncertain Moving Objects

 A moving object m () is a partial function and can have definition gaps
 If the maximum velocity of m is known, the uncertainty region of m can be
computed as a space-time prism.

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 93
Spatiotemporal Data Modeling

Predictive Moving Objects (I)

 Many applications are interested in a forecast of the (near) future
temporal evolution of a spatial object (e
(e.g.,
g hurricane research
research,
meteorology, fire management, disaster management, disease control)
 Example: Hurricane Ernesto (August 2012)
 Prediction is application-specific and afflicted with uncertainty
 Geometric aspect can be represented by spatiotemporal data types ()

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 94
Spatiotemporal Data Modeling

Predictive Moving Objects (II)

 Uncertainty aspect represented by a confidence distribution function that
is assigned an domain/application
domain/application-specific
specific distribution function
 Nature of these functions: probabilistic, fuzzy, possibilistic, rough, …
 Balloon metaphor taken from hurricane research
 Past movement of a hurricane resembles the string of a balloon
 Future prediction of a hurricane resembles the body of a balloon
 Balloon object = historical moving object + predictive moving object
(spatiotemporal snapshot)

 Observation:
Ob ti di (historical
dim(hi t i l moving
i object)
bj t)
≤ dim(predictive moving object)

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 95
Spatiotemporal Data Modeling

Predictive Moving Objects (III)

 Moving balloon objects represent the dynamics of moving objects as a
continuous sequence of balloon objects
 Past movement is updated (historical accumulation)
 New ppredictions and employed
p y pprediction models are p
possible due to new
insights and input factors
 Quality of predictions can be checked

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 96
Spatiotemporal Data Modeling

Spatiotemporal Partitions and Moving Objects in

Spatial Networks
 Spatiotemporal partitions (ongoing research)
 represent the temporal evolution of spatial partitions
 Maintain partition constraints over time
 Examples: change of temperature zones (weather maps on TV),
t
temporall evolution
l ti off iicing,
i d
development
l t off unemployment
l t rate
t
 Moving objects in spatial networks (ongoing research)
 Moving objects are spatially constrained
 Movement of spatial objects depends on speed, slope, number of
lanes, other moving objects, and so on
 Examples: car navigation,
navigation location
location-based
based services
services, mobile computing
computing,
traffic jams, pipelines (water, oil, power), Internet
 Continuous range queries interesting for car navigation
Example: Permanently report the three nearest restaurants while
driving
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 97
Spatiotemporal Data Modeling

Spatiotemporal Operations (I)

 Operations for projection to domain/range
 The
Th operation
ti deftime
d fti returns
t the
th times
ti for
f which
hi h a movingi object
bj t
(seen as a function) is defined.
deftime : ()  periods [periods  type for interval sets]
 The operation rangevalues returns the values assumed over time as a
set of intervals.
rangevalues : ()  range() [1D]
 The operation locations determines those components of the projection
of a discretely moving point into the plane that form a point object.
locations : (point)  point
 The operation trajectory determines those components of the projection
of a continously moving point into the plane that form a line object.
object
trajectory : (point)  line
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 98
Spatiotemporal Data Modeling

Spatiotemporal Operations (II)

 Operations for projection to domain/range (continued)
 The
Th operationti routes
t returns
t the
th projection
j ti off a discretely
di t l moving
i liline
object.
routes : (line)  line

 The operation traversed returns the projection of a continuously moving

line object.
t
traversedd : ()
( )  region
i f   {line
for li , region
i }

 The operation inst yields the time component of a (time, value) pair.
i t : intime
inst ( )  time
i ti () ti

 The operation val yields the value component of a (time, value) pair.
vall : intime
i ti ( )
() [intime
i ti = time
ti × ]]

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 99
Spatiotemporal Data Modeling

Spatiotemporal Operations (III)

 Operations for interaction with domain/range
 The
Th operation ti atinstant
ti t t restricts
t i t a moving
i object
bj t tto a given
i ti
time iinstant,
t t
resulting in a (time, value) pair (time slice operator).
atinstant : () × time  intime()
 The operation atperiods restricts a moving object to a given set of time
intervals.
atperiods : () × periods  ()
 The operations initial and final return the first and last (time, value) pair,
respectively.
initial, final : ()
( )  intime(())
 The operation present checks whether a moving object exists at a
given time instant, or is ever present during a given set of time
intervals,, respectively.
p y
present : () × time  bool present : () × periods  bool
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 100
Spatiotemporal Data Modeling

Spatiotemporal Operations (IV)

 Operations for interaction with domain/range (continued)
 The
Th operationti att restricts
t i t a moving
i object
bj t tto a value
l iin th
the range.
at : () ×   () [1D] at : () × range()  () [1D]
at : () × point  (point) [2D]
at : () ×   () [2D] with  = if dim()  dim() then  else 

 The operations atmin and atmax restrict a moving object to the times
when it is minimal or maximal with respect to the total order on this
space.
atmin, atmax : ()  () [1D]

 The operation passes checks whether a moving object ever assumed

(one of) the value(s) given as a second argument.
passes : () ×   bool with  = if dim()  dim() then  else 

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 101
Spatiotemporal Data Modeling

Spatiotemporal Operations (V)

 Derivable spatiotemporal operations by temporal lifting: automatically
derive spatiotemporal operations from spatial operations based on a clear
semantics

 Examples
 dist: point  region  real
Dist: (point)  (region)  (real)
Dist: hmpoint  hmregion  hmreal
 intersection: line  region  line
Intersection: (line)  (region)  (line)
Intersection: hmline  hmregion  hmline
 area: region  real
Area: (region)  (real)
Area: hmregion  hmreal

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 102
Spatiotemporal Data Modeling

Spatiotemporal Operations (VI)

 Let f : 1  …  n   be a spatial operation with i  {point, line, region}
and   {point, line, region, int, real, bool}.
} Its corresponding temporally
lifted version, that is, its corresponding spatiotemporal operation is
defined by
f : (
( 1)  …  (
( n)  ()
()
with
f(S1, …, Sn) :=
: {(t, f(S1(t), …, Sn(t))) | t  time}

 Examples
 Dist = dist
 Intersection = intersection
 Area = area

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 103
Spatiotemporal Data Modeling

Spatiotemporal Predicates (I)

 Problem: Temporal lifting of spatial predicates does not lead to predicates

 Example:
 Spatial predicate: inside : region  region  bool
 Temporally lifted spatial predicate:
inside : (region)  (region)  (bool)

 A lifted spatial predicate yields a defined value only on the intersection of

the domains of two moving objects.

 A spatiotemporal predicate is a function of type ()  ()  bool for , 

 {point, line, region}.

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 104
Spatiotemporal Data Modeling

Spatiotemporal Predicates (II)

 Basic spatiotemporal predicates
 defined
d fi d bby ttemporall lifti
lifting and
d ttemporall aggregation
ti
 each predicate has an own “preferred” temporal aggregation
 Examples
 disjoint: disjointedness required only on
the common lifetime
 inside: containment required
q only
y with
respect to the lifetime of the first operand
object
p
 Define default expected aggregation
gg g
behavior

 A spatiotemporal
p p p
predicate describes the continuous development
p of the
topological relationships of two moving objects over time.

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 105
Spatiotemporal Data Modeling

Spatiotemporal Predicates (III)

 Example: air plane (red) crosses a hurricane (blue)

time

t4 disjoint at t4
Disjoint during ]t3, t4[
t3 meet at t3

Inside during ]t2, t3[

t2 meet at t2
y
Di j i t during
Disjoint d i ]t1, t2[

t1 disjoint at t1
x

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 106
Spatiotemporal Data Modeling

Spatiotemporal Predicates (IV)

 Let P be a basic spatiotemporal predicate, S1 and S2 be two moving
objects and I be a (half
objects, (half-)) open or closed time interval
interval. Then predicate
constriction is defined as
PI(S1, S2) := P(S1|I, S2|I)
 Example (plane F, hurricane H)
 Inside(F, H) = false
 Inside|]t2, t3[(F, H) = true
 Two classes of topological predicates
 Instant predicates can be true for an instant time (but also for a period
of time)
equal, meets, covers, coveredBy
 Period predicates can only hold for a period of time
disjoint, overlap, inside, contains
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 107
Spatiotemporal Data Modeling

Spatiotemporal Predicates (V)

 A spatiotemporal predicate is an alternating temporal sequence of
topological relationships that hold over time intervals or at time points
points.

 A spatiotemporal predicate is an alternating temporal sequence of

topological
opo og ca relationships
e a o s ps a and
d bas
basic
c spa
spatiotemporal
o e po a ppredicates.
ed ca es

 Symbol “>>” is temporal composition operator

 Examples of spatiotemporal predicates for a moving point object and a

moving region object
 Enter := Disjoint >> meet >> Inside
 Cross := Disjoint >> meet >> Inside >> meet >> Disjoint
 Leave := rev(Enter)
 Cross ::= Enter >> Leave
 TempLeave := (Meet >> Disjoint)* >> Meet
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 108
Spatiotemporal Data Modeling

Spatiotemporal Predicates (VI)

 Examples of spatiotemporal predicates for two moving region objects

 Touch := Disjoint >> meet >> Disjoint

 Snap := Disjoint >> Meet

 Release := Meet >> Disjoint

 Bypass
yp := Snap
p >> Release

 Excurse := Meet >> Disjoint >> Meet

 Into ::= meet >> Overlap >> coveredBy

 OutOf := rev(Into)

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 109
Spatiotemporal Data Modeling

Spatiotemporal Predicates (VII)

 Examples of spatiotemporal predicates for two moving region objects
(continued)

 Enter := Disjoint >> Into >> Inside

 Leave
L := rev(Enter
E t )

 Cross := Enter >> Leave

 Melt := Disjoint >> meet >> Overlap >> Equal

 Separate := rev(Melt)

 Spring := equal >> Overlap >> meet >> Disjoint

 Graze := Disjoint >> meet >> Overlap >> (CoveredBy >> Overlap)* >>
meet >> Disjoint

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 110
Spatiotemporal Data Modeling

Spatiotemporal Querying (I)

 Tables
 CREATE TABLE Flight(id
Fli ht(id varchar(30),
h (30) route
t hmpoint);
h i t)
 CREATE TABLE Weather(name varchar(30), kind varchar(50),
extent hmregion);

 Q1: What are the departure and arrival times of flight LH 257, and how
long is the part of the route of this flight that lies within France?
CREATE TABLE France AS
(SELECT * FROM States WHERE sname = ‘France’);
SELECT min(deftime(route)) AS departureTime,
max(deftime(route)) AS arrivalTime,
length(trajectory(Intersection(route, ^territory))) AS lenInFr
FROM Flight, France
WHERE id = ‘LH 257’;

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 111
Spatiotemporal Data Modeling

Spatiotemporal Querying (II)

 Q2: When and where did flight LH 257 enter the territory of France?
SELECT i iti l( t( t territory))
initial(at(route, t it )) AS entry,
t
inst(entry) AS entryTime,
val(entry) AS entryLocation,
FROM Flight France
Flight,
WHERE id = ‘LH 257’;

 Q3: Determine the time periods when snow storm ‘Lizzy’

Lizzy consisted of
exactly three separate areas.
CREATE TABLE Lizzy AS
(SELECT extent
FROM weather
WHERE name = “Lizzy” and kind = “snow storm”);
SELECT deftime(at(noOfComps(extent) = 3,
3 TRUE))
FROM Lizzy;
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 112
Spatiotemporal Data Modeling

Spatiotemporal Querying (III)

 Q4: Where was United Airlines flight 207 at time 8:00 am?
SELECT val(atinstant(route,
l( ti t t( t 8:00))
8 00)) AS locAtEight
l AtEi ht
FROM Flight
WHERE id = ‘UA 207’;

 Q5: Where was United Airlines flight 207 between 7:00 am and 9:00 am?
SELECT trajectory(atperiods(route, 7:00..9:00)) AS loc
FROM Flight
WHERE id = ‘UA 207’;

 Q6: Which planes ran into a hurricane?

SELECT id
FROM Flight, Weather
WHERE kind = ‘hurricane’
hurricane AND
route Disjoint >> meet >> Inside extent;

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 113
Spatiotemporal Data Modeling

Spatiotemporal Querying (IV)

 DEFINE Enters AS Disjoint >> meet >> Inside;

 Q7: Which planes ran into a hurricane (reformulation)?

SELECT id
FROM Flight Weather
Flight,
WHERE kind = ‘hurricane’ AND
route Enters extent;

 Other definitions of complex spatiotemporal predicates

DEFINE Leaves AS rev(Enters);
DEFINE Crosses AS Enters >> Leaves;
DEFINE Bypass AS Disjoint >> Meet >> Disjoint;

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 114
Spatiotemporal Data Modeling

Spatiotemporal Querying (V)

 Q8: What is the number of planes that were entering snow storms or fog
areas?
SELECT count(*) as num
FROM Flight, Weather
WHERE kind = ‘snow storm’ OR kind = ‘fog’
GROUP BY route Enters extent;

(
 CREATE TABLE Bird(swarm varchar(40),
( ) movement hmpoint);
p )

 Q9: Which swarms fly together, then take different routes for some time,
y meet again?
and finally g
DEFINE Remeets AS _ >> Meet >> Disjoint >> Meet >> _;
SELECT A.swarm, B.swarm
FROM Bi d AS A
Bird A, Bi
Bird
d AS B
WHERE A.movement Remeets B.movement;
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 115
Outline

Outline
1. Introduction
2 Spatial
2. S ti l D
Datat M
Modeling
d li
3. Spatiotemporal Data Modeling
4. Open Research Topics

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 116
Outline

Outline – Open Research Topics

 Topics not covered in this tutorial
 Research
R h challenges
h ll

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 117
Open Research Topics

Topics Not Covered in this Tutorial

 Goal of the tutorial: Learn concepts first
 Data structure design for spatial and spatiotemporal data types
 Geometric algorithm design for spatial and spatiotemporal operations and
predicates (require methods from Computational Geometry)
 Efficient implementation techniques for spatial and spatiotemporal joins
 Spatial
p and spatiotemporal
p p q
query
ypprocessing
g
 Spatial and spatiotemporal indexing

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 118
Open Research Topics

Research Challenges (I)

 Spatial partitions and spatial networks as first class citizens in spatial
databases (Map Algebra)

 Query languages for spatial partitions and spatial networks

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 119
Open Research Topics

Research Challenges (II)

 Vague and fuzzy spatial objects

 Some models for fuzzy spatial objects available (especially from GIS field)
 Problem: Implementation of fuzzy spatial objects
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 120
Open Research Topics

Research Challenges (III)

 Three-dimensional spatial objects

 Applications: Meteorology, GIS, soil science, earth science

 Problem: Computational Geometry is immature with respect to three-
dimensional data handling
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 121
Open Research Topics

Research Challenges (IV)

 Moving objects (3D + time)

 Applications: Meteorology, hurricane research, tsunami research,

earthquake research, volcano research
 Construction of moving objects from gridded sensor data (icing problem)
Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 122
Open Research Topics

Research Challenges (V)

 STU objects (Space + Time + Uncertainty)

 Applications: Modeling of natural phenomena in general, geosciences,

hurricane research, meteorology

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 123
 Thank You
for Your Attention
and Interest!

Spatial and Moving Objects Databases: State of the Art and Future Research Challenges 124