Pervasive Computing and Space: A
Knowledge Based Approach
M. Palmonari
University of Milan-Bicocca
Department of Computer Science,
Systems and Communication
Artificial Intelligence Lab
Outline

Motivation




Correlation of Traffic Alarms on a Highway




From 1D to 2D, with different spatial relations
The Hybrid Logic Perspective
A Case Study: correlation of events for the Supercentro
Project



Correlation of alarms for the system SAMOT
Monodimensional Space, Modal Logic Approach
Commonsense Spatial Model and Commonsense Spatial
Hybrid Logic


Pervasive computing and information correlation
Correlation of information with (qualitative) spatial reasoning
A Knowledge Representation approach: a strategy focusing on
space
RoadNetwork Spatial Environment
Extending the Commonsense Spatial Hybrid Logic
Few Concluding Remarks (KR)
Outline

Motivation




Correlation of Traffic Alarms on an Highway




From 1D to 2D, with different spatial relations
The Hybrid Logic Perspective
A Case Study: correlation of events for the Supercentro
Project



Correlation of alarms in the system SAMOT
Monodimensional Space, Modal Approach
Commonsense Spatial Model and Commonsense Spatial
Hybrid Logic


Pervasive computing and information correlations
Correlation of information with (qualitative) spatial reasoning
A Knowledge Representation approach: a strategy focusing on
space
RoadNetwork Spatial Environment
Extending the Commonsense Spatial Hybrid Logic
Few Concluding Remarks (KR)
Pervasive/Ubiquitous Computing
Ubiquitous/Pervasive Computing: a new way of conceiving the
interaction among humans (users) and computing devices.
“Computers are disappearing” and computational power can be
embodied in every object populating the environment
(ubiquitous)
Technologies
sensors, PDAs, mobile phones,
wi-fi, GIS, and so on...
From an architectural perspective: distributed systems with
components that are mobile and embedded in the environment
Ubiquitous: ubiquitous access to information
Pervasive: pervasive acquisition and processing of information
Interpretation of information

Pervasive Computing:


huge number of information acquisition devices
distributed into space
support users in their daily tasks / in special tasks
(e.g. context aware applications for PDA vs.
Control&Monitoring Systems)

Interpretation of information (2 meanings):

Technologies enable to collect data to study specific
phenomena (e.g. sociological, environment, economical)


GIS, mobile GIS, GPS, …
Applications that need to exploit information for their
purposes (e.g. Control and Monitoring Systems, Context
Aware applications, Location Based Services)
Correlation of information

Problem: information provided by acquisition
devices on the environment may loss significance
as a huge number of sensors tend to produce a
overload of information

Information correlation to integrate different
sensors data



Integration of information from heterogeneous data
sources for decision support systems
Context Awareness
Knowledge based approach:


Integration into a higher level non knowledge based
techniques
Generality
A knowledge based approach

Usually: information fusion

Statistic based techniques:




Very efficient
Domain specific
Data for calibration? Training data sets?
A Knowledge based approach



A knowledge based correlation level
Integrating non knowledge based techniques into a
knowledge based level
Background:


Control&Monitoring systems: experience from experts
systems
Context Awareness: attention to high level/general/formal
KR techniques (e.g. Ontologies)
A 4-level conceptual architecture
4 level Architecture
presentation
actions
dissemination
…
ACTUATION
CORRELATION
LOCAL
INTERPRETATION
ACQUISITION
environment
4 level Architecture
presentation
actions
dissemination
…
ACTUATION
scenarios
CORRELATION
events
LOCAL
INTERPRETATION
data
ACQUISITION
environment
The 4-level Architecture
presentation
actions
dissemination
…
ACTUATION
KB
scenarios
CORRELATION
events
LOCAL
INTERPRETATION
data
ACQUISITION
environment
KB correlation of information as spatial
reasoning

Space/time is the homogeneous ground for correlation of
heterogeneous information



KB correlation level:




Events: information with spatial location and temporal
duration (explicit vs. implicit)
Scenarios: specification of significant logical correlations
among local descriptions, with focus on spatial / temporal
correlation (spatial | temporal | spatiotemporal scenarios)
Generality (heterogeneous information  events/scenario)
Knowledge Models closer to human experience
Scalability/reusability
Focus on space/time: correlation of information as spatiotemporal reasoning


Correlation of localized events
Focus on space (computational concerns)
A KR approach: the strategy
A general strategy:
1. Define a spatial model


2.
Choose a suitable formal language to reason
about information through the spatial model


3.
Which kind of model?
Qualitative vs. quantitative
A logic!
Expressiveness vs. Tractability
Exploit the language to define correlation
axioms

How to derive interesting scenarios from logical
correlation of events
A KR approach: Qualitative Spatial
Reasoning

Qualitative Spatial Reasoning






Quantitative spatial information is often imprecise or
unavailable (e.g. localization into a room)
Meaningful spatial information is often related to
qualitative/semantic representation (e.g. to be into Central
Park vs. geodetic coordinates)
Reasoning with quantitative models is generally
computationally intractable
Closer to human (users) representation
From different research areas: KR, GIS, ontology, cognitive
sciences, robotics
Focus on relational aspects






Topological relations (connection, parthood, …) - RCC calculus,
mereotopological theories
Direction relations (cardinal directions)
Inclusion/containment relations
Distance relations (far/near)
Fuzzy relations
Mental maps - Kuipers
A KR approach: Modal Like Logics

Qualitative Spatial Models: relational models,
spatial entities and spatial relations (calculus:
algebra vs. logic)

Modal Like Logics: fragments of First Order Logic
representing relational aspects with a particular
perspective over reasoning





Semantics: relational structures (or topological spaces)
Generally: good computational properties
Compact formulas and proofs
Local perspective on reasoning (formulas have context
dependent meaning)
We will see them in action!
From monodimensional to
bidimensional space and maps
Walking through different classes of spatial models:
1.
Correlation of events on a highway



2.
Different spatial relations: Commonsense Spatial Hybrid
Logic (CSHL)



3.
Correlation Module of SAMOT
Spatio-temporal correlation (1-d space-time)
Adaptation of temporal modal logic
Examples from Smart Home applications
Places and Commonsense Spatial Relations on 2D
Hybrid Logic, Calculus, Proofs
Correlation of events on a Road Network



Event Correlation Model for Supercentro
Modification of CSHL to represent the road network and other
relevant spatial entities (2D)
Examples of complex scenarios definable in the logic
Alarm Correlation with a
Spatio Temporal Modal Logic
SAMOT a System for Monitoring Traffic Flows on an Highway
The SAMOT project


To deliver a monitoring and control system to
support traffic operators of Italian highways in their
activities to improve traffic safety, congestion
prevention and effectively act in case of
emergencies
SAMOT (System for Automatic MOnitoring of Traffic)





Automatic detection of traffic flow anomalies
(queue, stopped vehicle, wrong-way driving vehicle…)
Automatic control of some traffic flow anomalies
Alert traffic operators in case of emergencies
Correlation of local sensor states into complex
anomalous traffic patterns (Detection Rate
improvement)
Filtering of alarms according to highway sections’
peculiarities (False Alarm Rate reduction)
(1) Monitoring Agencies: set of videocameras, each one devoted to the
monitoring of a small portions of the highway
1
2
4
3
(1) Monitoring Agencies: set of videocameras, each one devoted to the
monitoring of a small portions of the highway
1
2
4
3
(2) A Video Image Processing board is
associated to every camera, in order to
interpret local traffic condition (e.g.
slowTraffic, queue, wrongWayDrivingVehicle,
stoppedVehicle)
(1) Monitoring Agencies: set of videocameras, each one devoted to the
monitoring of a small portions of the highway
1
2
4
3
(2) A Video Image Processing board is
associated to every camera, in order to
interpret local traffic condition (e.g.
slowTraffic, queue, wrongWayDrivingVehicle,
stoppedVehicle)
(3a) Warning messages are generated
(appropriate message and operating mode,
e.g. duration time) and automatically shown
on Variable Message Panels (VMP)
(3b) support to operators: GUI for data
visualization, remote device control and
diagnosis, user profile management (security
settings), system configuration (type and
number of devices, relation between alarms
and control actions)
(3c) system functioning: e.g. management
of video flows (default sequences can be
retrieved and modified), image management
(4) Correlation of Local Traffic States into
Anomalous Traffic Patterns  suggestions on
Control Actions directed to traffic operators
to support anomalous traffic management
Knowledge Representation Approach to
Alarm Correlation …
VIPs’ outputs + system info
(interpretation on a local traffic situation at a given time,
e.g. a queue at time t in region XX )
events
anomalous traffic patterns
(e.g. a stopped vehicle “followed” by a queue 
maybe(incident)  warning message to operators)
spatio-temporal scenarios
-
-
ST
-
12:00 AM
12:05 AM
12:10 AM
-
Q
ST
ST
-
Q
ST
ST
-
ST
Q
ST
-
-
ST
ST
-
ST
-
-
• Road section representation as sequence of cameras
• Traffic anomalies detected by VIPs (referring to adjacent cameras) are
correlated and interpreted as possibly anomalous traffic pattern
Knowledge Representation
Approach to Alarm Correlation …
based on traffic operators’ experiential knowledge about spatial and
temporal adjacency relationships between VIPs’ interpretations  set
of significant anomalous traffic patterns
(1) Extension of anomalous traffic patterns
ti
ti+1
-
Q
Q
-
-
-
-
-
Q
Q
ST
ST
-
-
Queue
(2) Reductions
(3) Shifting
ti
ti+1
Q
ST
ST
-
-
ST
ti
ti+1
-
Q
-
ti
ti+1
Slow traffic
Q
Q
Q
-
-
-
-
-
Q
Q
-
-
-
-
Q
Q
Q
-
-
-
Q
Q
-
ST
Q
ST
-
Q
Q
ST
ST
Q
ST
(4) Composition of patterns
Towards a formal model of the MCA
Key elements:
SPACE …
…and TIME
Exploit
adjacency
relations
along both
time and
space.
time
space
space
with MODAL LOGIC: the logic
for relational structures!
1 - The spatio-temporal model
succession of time points
Space: space intervals along a discrete succesion
Time
Time: time intervals defined over a discrete
of adjacent atomic intervals
Spatial-Temporal Region (STR):Arbitrary
quadrilateral shapes defined by a time interval
and a space one.
Stripe Region (SR): STRs with a minimum
dimension along Space or Time.
Atomic Region : a minimum dimension along
both space and time.
Space
Classic Kripke semantics: each STRegion is a possible world…
Formally the truth is defined as usual for
modal logic
intuitively
ST<ti,tj,sh,sk> ⊨ φ
φ is true in the ST region defined by the time interval <ti,tj>
and the space interval <sh,sk >
2 - The ST modal language:
Atomic Alarms as propositional variables
Local alarms are interpreted as propositional constants true
If
alarm ST-regions
is
atno
atomic
identified by one section of highway
produced
by
and a minimal
time interval.
VIPs, “Normal
Traffic” (NT) is
assumed to be
true
Propositional
constants to
represent local
alarms:
-
-
ST
Q
Q
ST
Space
Queue (Q)*
Slow Traffic (ST)*
Stopped Vehicle (SV)*
Wrong Way Vehicle (WV)*
* = Anomalous Traffic
Situations (ATS)
-
Anomalous traffic patterns
Alarm correlation is achieved combining modal
operators in formulas of the ST-logic language.
T0
T1
correlate
-
-
Creation
-
Modal operators allow to
facts- true -at Q
different time in different sections of the highway
Q
T0
exploiting adjacency relations along
both
time
andQ
T1
Q
Q
space.
T0
T1
-
-
-
-
-
-
Q
ST
Extension
-
-
-
-
ST
ST
-
-
Decomposition
-
Q
Q
ST
Q
Q
ST
-
Q
Q
-
ST
Q
-
Anomalous traffic patterns (ATPs) are represented
by propositional constants trueT0
on stripeQ regions
ST
ST
along space.
T1
ST
Reduction
Q
Q
Q
-
-
-
-
-
Shifting
T0
T1
-
Q
Q
Q
-
-
-
-
-
Q
Q
Q
-
-
Composition
T0
T1
-
Q
Q
-
ST
Q
ST
-
Q
Q
ST
ST
Q
ST
2 - The ST modal language:
modal operators for spatio-temporal relations
W : within
AS, AT, A : after
BS, BT, B : beginning
ES, ET, E : ending
+ the respective duals [X]
+ the transpose -A, -B, -E
Time
Modal operators
Current
ST-Region
[[FS]], [[FT]], [[F]] : first
[[LS]], [[LT]], [[L]] : last
E.g. for the after relations:
AS
:
Space
 holds at some ST-region  spatially beginning immediately after the spatial end of a current
one.
AT
:
 holds at some ST-region  temporally beginning immediately after the temporal end of a
current one.
A
:
 holds at some ST-region  beginning immediately after the spatial and temporal end of a
current one.
2 - The ST modal language
Where can we move?
Where
can we
move?
Within: into some region within the current one W
Time
Modal operators can be viewed as
ways to move into the regions in
which the satisfiability of the formula
after the operator must be checked
Space
2 - The ST modal language
Into some
region
spatially
beginning…
Time
Where can we move?
BS
…or
temporally
ending the
current one
ET
Space
2 - The ST modal language
Where can we move?
Into some region before the current one along space, time or both… use the
transposes of after!
Time
-AS -A -AT
Space
2 - The ST modal language
Where can we move?
[[FS]]AS
Into the spatially
last region of
some region
spatially after the
current one
Space
3 – Correlation Formulas:
anomalous traffic patterns with ST-Logic
Time
Some example of formulas
used for correlation… how
to infer anomalous traffic
patterns…
ATS ∧
-ATNT
→ CREATION
Q
ST
ST
Q
Q
NT ∧ -ATATS → DELETION
Space
ATS ∧ -ATWATS ∧ (BS CREATION ∨ ESCREATION) → EXTENSION
[[LS]]CREATION ∧ [[FS]]DELATION → SHIFTFW
[[LS]]DELATION ∧ [[FS]]CREATION → SHIFTBK
3 - Correlation Fomulas
the set of ST-formulas
Implementation and Limits

Implementation into a rule based system


Limits & further works:





Model vs. Implementation
Adaptation of interval based temporal logic
ST-Logic is undecidable (the problem of spatio-temporal
representation)
Spatial representation too simple for most application
domains (1-D)
Heterogeneous events
From 1D to 2D spatial representations
A Commonsense Spatial
Hybrid Logic
For information correlation in pervasive computing
A pervasive computing example:
a smart environment
Consider a smart environment providing a sensor platform installed in
a building in order to monitor a significant portion of it…
Different sensors (e.g. a camera, a
fire detector, a broken-glass sensor)
Sensors can return local
descriptions (e.g. detection of a
person, alert for a broken glass)
The environment: significant
location are the different rooms
(each with specific properties), or,
eventually particular portion of
them
A correlation example: neither a broken glass nor a person detected by
the camera are per se a proof of intrusion, but those two facts considered
together may lead to infer that a stranger is entered into the house
passing through the window and walking in the corridor
Knowledge Representation approach
Spatial model
What does IN mean?
Semantics/model
Language
Spatial concepts
e.g. L.int.ar is in U7
I’m in L.int.ar (here)
U7 is between U6 and U2
Reasoning/inference
Spatial inference
e.g. from L.int.ar is in U7
and my PDA is in L.int.ar
infer: my PDA is in U7
1 - Commonsense Spatial Model: basic
notions
Commonsense Spatial Model (CSM)
…a relational approach:

Focus on
relevant entities of the
environment

PLACES as aggregates of
information
e.g. rooms, buildings, devices,
…

relevant spatial relations
among places

COMMONSENSE SPATIAL
RELATIONS (CSR)
e.g. in, north of, adjacent to, …
A relational structure CSM = P , R s
P is finite, Rs binary CSRs
1 - Interesting classes of Commonsense
Spatial relations


CSM is very weak, but…
Interesting classes of CS relations grouped according to
formal properties
Different Relations classes:
Connection
Containment
Orientation
1 - Interesting classes of Commonsense
Spatial relations


CSM is very weak, but…
Interesting classes of CS relations grouped according to
formal properties
Different Relations classes:
Connection
The basic graph structure:
reachability of one place
from another
Containment
Orientation
1 - Interesting classes of Commonsense
Spatial relations


CSM is very weak, but…
Interesting classes of CS relations grouped according to
formal properties
Different Relations classes:
Connection
Containment
Qualitative, semantically
qualified location, but also
hierarchically order places
of different kind (e.g. a PDA
in a room)
Orientation
1 - Interesting classes of Commonsense
Spatial relations


The notion of CSM is very weak, but…
Interesting classes of CS relations grouped according to
formal properties
Different Relations classes:
Connection
Containment
Orientation
Qualitatively ordering
places in 2D/3D, with
respect to reference points
2 – the Logic.
CSMs as Semantics for a Multi-Modal Hybrid Language
 Modal Logic allow to reason over relational structures in a very intuitive way.
 Every CSM is a relational structure that can be taken as the semantic
specification for a multi-modal hybrid language.
Our basic CSM multi modal language
Propositional variables
Modal operators
Logical connectives
→ , ¬ ,∧ ,∨
p, q, …
Semantic: M = (F CSM , V
)
◊P , ◊IN , ◊N , ◊E , ◊S , ◊W + □P , …
A model =
a basic CSM + valutation
Modal operators allow to explore the spatial model to check
satisfiability of formulas from a current place (M,w ⊨ φ ?)
Class of CSRs provide the meaning to modal operators according to
their formal properties
2 – The Logic
Expressivity: CS Modal formulas (diamond)
φ
∨
φ
∨
φ
RIN
RIN
Modal logic:
diamond/boxes
operators
RIN
⊨ ◊IN φ
◊P , ◊IN , ◊N , ◊W ,…
Modal Local Perspective:
from a current place…
e.g. with a containment
operator ◊IN
2 – The Logic
Expressivity: CS Modal formulas (box)
Modal logic:
diamond/box
operators
⊨□Pφ
φ
RP
∧ R
P
φ
Modal Local Perspective:
from a current place…
□P , □ IN , □ N , □ W ,…
e.g. with a containment
operator □ P
2 - The Logic
Modal Logic is not enough: Hybrid Logic
smoke_sensor
⊨ @livingroomφ
kitchen
livingroom
φ

Hybrid Logic empowers modal expressivity adding


NOMINALS: names for places ([i,j,…]  kitchen, PDA21, …)
SATISFACTION OPERATORS: @i operators to move to specific
places ([@i]  @livingroom,...)
2 - The Logic
Axioms for standard CSMs

How to specify the meaning of spatial modal
operators in proof theory?


Meaning of operators is related to properties of the
respective accessibility relations
Stantard CSMs

Relations are characterized by formal properties and
grouped together in classes

Definition enough precise for axiomatixation

Definition of a calculus for Commonsense Spatial
Reasoning with tableaux
Axioms for standard CSMs
Local & Global truth
Is there a bar/restaurant in this place?
Here truth is context dependent… it depends on where
the formula is evaluated
Is the bar cheap?
The truth of the formula depends only on what is true at the
place denoted by “bar”
CS Hybrid Logic for pervasive systems

Local and global perspective



Local, context dependent reasoning exploiting
modal formulas
Reference to specific places by hybrid
formulas
Flexibility (frame definability)


Easy to add extra operators and characterize
them with pure formulas
Easy to adapt calculi in different contexts
The Tableau-based Calculus
Reasoning with CSM 2: tableaux based
proofs (1)
3 – Correlation with CS Logic
Reasoning with CSM 2: tableaux based proofs (2)
A very simple case of alarm
correlation… propagation of alarm
with filter:
“If there are two alarms inside a
room we can say that in that room
there is an alarm” (one is not
sufficient)
Now, knowing that two sensors of the kitchen detected an alarm
infer that in the kitchen there is an alarm
Proof by refutation of
Reasoning with CSM 2: tableaux based
proofs (2)
Discussion: what form of reasoning do we want?
By refutation? Backward? Forward? Production rules?
All logics?
Reasoning… but how?
Different approaches with different {power, complexity,
correctness, perspective, applicability}

Pure (Hybrid) Logical Calculus


Model checking (what is true in a model?)


Weak, but it is possible to make the Model Checking stronger
+ some spatial inference
Same model with different logical languages


E.g. Tableaux based calculus (defined but not implemented)
E.g. into a Logic Programming Framework like AnsProlog
Approximated by other techniques

E.g. by a production rule system (JESS)
The Supercentro Project.
Correlating information on a road
network spatial environment
Pervasive Technologies in Milan

MILAN: lot of sensors installed in the environment

Information about:






Sensors organized into subsystems
Different actions can be taken on the basis of such information

Information diffusion technologies


Broadcasting, web
Control actions



Traffic (magnetic loops, CCTV)
Air Pollution and other environmental condition
Violations
Events occupying roadbed
Automatic (Urban Traffic Control systems [UTC], Visual message panels
[VMPs])
Support operators’ decisions
Sensors and devices managed by different subsystems: need for
integration of heterogeneous information and coordination of
actuation strategies and actions
Supercentro: the project


Supercentro is an ongoing project carried out by Project Automation
S.p.A. for the development of a platform integrating different information
sources producing and storing information about phenomena related to
mobility (or relevant to it) in the City of Milan.
Goals:

support qualified operators in monitoring such phenomena in
order to take suitable actions



diffuse relevant information to citizens


Web, radio (RDS/TMC), mobile
select retroactive actions autonomously





UTC management
…
UTC coordination
VMPS
CCTV
…
Higher level interpretation of data integrating different
sources
Decisione
Correlation
Monitoraggio
e Controllo
Integrazione Produttori Di Informazioni
Local Interpretation
Controllori
Locali
Controllori
Locali
Controllori
Locali
Controllori
Locali
Centrali di
Monitoraggio e
Controllo
Pannelli VMS
Sensori Meteo/Aria/Rumore
Acquisizione
e Attuazione
I° Livello
Centrale Gestione Traffico
Elaborazione
Presentazione
Centrale Informazioni Traffico
Acquisizione
Dati
II° Livello
Diffusione
Aree
Controllate
Sensori Flussi di Traffico
TVCC
Rilevamento Infrazioni
Regolatori Semaforici
The Supercentro System
Mappa di Milano con la rappresentazione degli impianti installati sul territorio
(Livello Installazione Periferica). Si distinguono installazioni semaforiche (1),
installazioni TVCC (2), installazioni periferiche Pannelli a Messaggio Variabile (3),
installazioni periferiche di conteggio e classificazione del traffico (4)
The Supercentro System
Dettaglio delle misure in tempo reale delle tipologie di transiti passati
attraverso varco ZTL (transiti totali, validi, non validi, sospetti, …)
The Supercentro System
Mappa di Milano con la rappresentazione degli impianti installati sul territorio
(Livello Dispositivo Periferico). Si distinguono regolatori semaforici (1), telecamere
(2), pannelli a messaggio variabile (3), sezioni di misura del traffico (4)
The Supercentro System
Dettaglio delle misure (24h precedenti) di alcuni inquinanti
misurati da stazioni di rilevamento fisse
The Supercentro System
Dettaglio delle misure in tempo reale (ultimi dodici valori con periodo di
campionamento 5min) della sezione di misura di “Via Canova direzione Via Pagano”,
sono rappresentate la velocità media, il flusso totale e il tasso d’occupazione.
The Supercentro System
Rappresentazione dello stato del traffico in modalità “colorazione archi”, in
particolare arco misurato (possiede almeno una sezione di misura in relazione con
l’arco) nello stato Fluido – Azzurro. Nel tooltip il dettaglio dei dati misurati.
The Supercentro System
Rappresentazione dello stato del traffico in modalità “colorazione archi”, in
particolare arco stimato (non possiede sezioni di misura in relazione con l’arco).
Nel tooltip il dettaglio dei dati stimati.
The Supercentro System
Rappresentazione dello stato del traffico in modalità “flussogramma colorato”, in
particolare arco misurato (possiede almeno una sezione di misura in relazione con
l’arco) nello stato Denso – Giallo. Nel tooltip il dettaglio dei dati misurati.
NB La modalità “flussogramma colorato” prevede che l’arco abbia uno spessore
proporzionale alla misura del flusso totale di veicoli mentre il colore rappresenta lo stato
(se determinabile).
The Supercentro System
Rappresentazione dello stato del traffico in modalità “flussogramma colorato”, in
particolare arco stimato (non possiede sezioni di misura in relazione con l’arco).
Nel tooltip il dettaglio dei dati stimati.
NB La modalità “flussogramma colorato” prevede che l’arco abbia uno spessore
proporzionale alla misura del flusso totale di veicoli mentre il colore rappresenta lo
stato (se determinabile).
The KB approach
KB approach:
 From first interpretation: events (elementary events)



traffic events, environmental pollution, road occupation, diagnostic
alarms, operative alarms (triggered by citizens/users)
To Correlation: infer scenarios from events
Actuation

Associating to scenarios:








the activation of specific inter-area traffic regulation plans;
the provision of information to citizens through VMPs;
the activation of specific sequences of camera views;
the activation of thematic presentation to assist operators;
suggestions of actions to take;
a high level analysis of data acquired by the subsystems in order
to study when some scenarios are more likely to occur;
combinations of the previous ones.
Configurability: provide a tool to enable operators to define
their correlations
The KB approach
KB approach:
 From first interpretation: events (elementary events)



traffic events, environmental pollution, road occupation, diagnostic
alarms, operative alarms (triggered by citizens/users)
To Correlation: infer scenarios from events
Actuation

Associating to scenarios:








the activation of specific inter-area traffic regulation plans;
the provision of information to citizens through VMPs;
the activation of specific sequences of camera views;
the activation of thematic presentation to assist operators;
suggestions of actions to take;
a high level analysis of data acquired by the subsystems in order
to study when some scenarios are more likely to occur;
combinations of the previous ones.
Configurability: provide a tool to enable operators to define
their correlations
The Spatial Representation approach
1.
Define a spatial model


2.
Which logic to reason over this model?


3.
Which spatial model to chose in order to define
interesting scenarios?
Starting from DB and event localization
Hybrid Logic?
How to modify CS Hybrid Logic?
Defining correlations


Exploit the logic to infer interesting scenarios
Which scenarios? How can be defined?
1 – The spatial environment
1 – The spatial model: entities

A Road Network representation: the GVPO

GVPO: main road network oriented graph




Events can be spatially referenced on the GVPO
Other regions overlapping the GVPO




Nodes: Intersection with traffic regulators
Edges: oriented road sections
Areas of interest (e.g. P.le Loreto Area)
Main traffic flows directions (MTFD)
…
Devices

Sensors and actuation devices located in the
environment
Four types of entities:
intersections (I) | edges (E) | regions (A) | devices (D)
1 – The spatial model: relations

Primitives

Spatial Connections among intersections and edges


Turn Connections among edges


(devices  GVPO  regions)
Direction relations


From Administrative Code
Containment relations


topological
Order among entities w.r.t. reference points (e.g. Trade
Fair, City Center)
Derivates

Origin/Destination, General Connections, Containment
Inverse, …
1 – The spatial model
Areas (regions)
Main traffic flows directions (regions)
1 – The spatial model
Directions
SubAreas
1 – The spatial model
General Connections
1 – The model. Formally…
How are those relations intensionally characterized?
Many from Commonsense Spatial Model (in, direction)…
2 – The Logic: RoadNetwork+ HL

The model is a relational structure



Hybrid Logic: modify CS Hybrid Logic (exploiting HL
properties)
 Different Connection operators
 β is diadic: φβψ means that being on an intersection (i),
φ it is true on an edge located between (i) and an
intersection on which ψ is true.




Four types of entities
Relations
<TC>: turn connection
φ
ψ
Typing (the four types of entities)
Different cross properties
Axiomatization

φβψ
Defining properties of relations within the language
2 – The Logic: RoadNetwork+ HL
φβψ
φ
<TC>φ
ψ
slowT β int
φ
SlowT
int
slowT
<TC>slowT
Axiomatization
3 – Correlation: scenarios definition
Formulas identify patterns of events
 Basics:

slowT β slowT β int
φ β ψ β int
slowT ∧ <TC>slowT
φ ∧ <TC> ψ
SlowT
SlowT
int
φ
ψ
int
slowT
slowT
φ
ψ
Types of scenarios
Three main types of general scenarios:

N-connection patterns



OutgoingScenarios


Homogeneous vs Heterogeneous
Linear vs. Tree-like
What is occurring in outgoing edges/areas from a given
edge/region?
DirectionScenarios

Exploiting directions to focus on specific edges/regions
3-connection patterns
e.g. 3-connection patterns
a)
Linear patterns
= φ1
= φ2
= φ3
b)
Tree-like patterns
Homogeneous patterns: φ1 = φ2 = φ3
Else: Heterogeneous patterns
3-connection Scenarios
= heavy congestion
= congestion
= dense
= fluid-dense
= fluid
= very fluid
3-connection
Scenario
3-connection Scenarios
Outgoing Scenarios
B
All outgoing edges
C
Some outgoing
MTFD
D
Some outgoing
areas
Outgoing Scenarios
C
Some outgoing
MTFD
Directed Scenarios
Direction
relation
“trade fear”
Modal operator
direction “trade
fear”
Constraint on
other scenarios
Directed Scenarios
Modal operator
direction “trade
fear”
Constraint on
other scenarios
Directed Scenarios
Direction
constraints on
Outgoing
Scenarios
Few Concluding Remarks (KR)



General strategy focusing on space
Different spatial models, different languages,
different representational issues
Hybrid Logic good perspective on reasoning in
these domains:




Modal Perspective on relational structures
Good expressive power
Flexibility in defining axiomatization and calculus
KR with logic: a good framework for models



Model vs. implementation
Control on assumptions with proofs
Generality
Questions?
Than you for the attention!
Descargar

Ubiquitous computing scenarios