System and Solutions for Mobile Pervasive Computing Environments Tutors: Prof. Romano Fantacci, Ing. Laura Pierucci, Prof. Imrich Chlamtac PhD final exam Firenze, 8 th April, 2009 David Tacconi
Outline
Introduction
System Architecture
Applications and challenges
Intelligent Transportation System (ITS) Application scenarios:
Routing issues
Data management issues
Mobile Social Network (MSN) application scenario:
Real implementation
Social analysis
Routing framework
Service evolution
Conclusions
Introduction
Mobile Computing:
the possibility of being able to use a computing device
even when being mobile and therefore changing location
Pervasive Computing:
a halo of embedded devices immersed into reality
able to provide information to a human or to another device
about the environments he is immersed in
Mobile Pervasive Computing:
Devices moving around
Looking for information from sensing devices
Not necessarily connected to a central server
Opportunistically exchanging information on the fly
System Architecture
Application Scenarios
ITS - scenario 1:
Mobile Nodes querying a large WSN (cars looking for a free parking)
With the absence of a central server
Need for new routing framework to handle a WSN with a mobile sink
ITS - scenario 2:
Mobile nodes querying gateway nodes
Exchanging information on the fly
Need for new data management techniques for handling large amount of volatile data
Application Scenarios
Mobile Social Networking
Nodes interact among them on the basis of users’ profiles and interests
Groups of friends get created in a localized way
In particular we deal with:
A real implementation for smartphone
Social analysis deriving from the use of this mobile service in a real environment
Routing algorithm based on degree of friendship
Service evolution driven by mobility
ITS: mobile sinks querying a large WSN
Mobile Sink querying a WSN
Representing cars looking for a parking place
System architecture specifically tailored for ITS scenarios:
MS e.g. a car
Sensing nodes e.g. presence sensors
Gateway nodes e.g interface nodes for the MS NOT providing continuous connection to the MS
ITS: the routing framework
Geographic forwarding
MS experiences frequent disconnections
Deadlock management
Mobility prediction
Load balancing strategies
Delay aware routing
Energy aware routing
ITS: simulation results
Simulations performed in Omnet++
Network dimensioning
Comparative study to evaluate load balancing techniques
Time to first node failure evaluated varying mobility and # of nodes
ITS: Data management issues
Data management and compression scheme :
Are needed in context aware applications for mobile nodes or in ITS applications
Largely deployed WSN
WSN can be considered as a Sensor Map (Image)
Local information has to be more precise while only coarse approximation can be kept for further information
Wavelet compression and data management scheme can help
Application scenario is described in figure
Data management for large WSN: Simulation results
Simulation conducted in Omnet++
Large number of sensors (128x128 sensor grid)
Variable number of nodes (100 – 600)
Variable mobility pattern
Distortion between real sensor image and stored sensor image is measured
Mobile Social Network Scenario: DTN MN5 MN1 MN2 MN4 Back Haul Connection MN5 MN3 MN 7 MN 6 Node Movement Opportunistic Information Exchange BlueTooth Module WiFi Module Mobile Node MN
State of the Art
Opportunistic networking has been deeply investigated from a theoretical point of view:
Bionets EU project
Haggle EU project
Several conferences and research intiatives
Only few real developments have been proposed:
To understand networking performance of the implemented protocols
To investigate social aspects related to proximity communications
Motivation for a real implementation
Mobile Phones are used:
For Voice/video calls
As Messaging device
As MP3 player
As Cameras
To surf the web
What if users could share data:
Using their mobile phones
Leveraging on proximity communications rather than relying on a backhaul connection
Simply editing their preference and search options every once in a while
Putting the mobile phone in the pocket and then TRANSPARENTLY exchanging information when meeting other users, according to personal interests
A middleware for pervasive environment
U-Hopper:
User centric Heterogeneous Opportunistic Middleware
provides users with ‘missing’ functionalities
Is Java + Bluetooth based
Can be used on every phone with J2ME support or Linux J2SE laptop
Supported applications (december 2008):
Profile editing (limited)
Advertisement and Business card exchange
Sensor data reading (images) and exchange
Contact exchange to trace contact evolution
Ring-tones exchange
U-Hopper System Architecture
Profile Manager (PM):
handles the creation/update/deletion of the user profile. Such profile can be explicitly created by the user, and dynamically updated on the basis of users daily activities .
Service Container (SC) :
is the environment where context-aware services are executed. Such Container provides seamless access to resources such as content storage, opportunistic data retrieval, etc.
Content Manager (CM):
stores permanently any data item considered as relevant to the Interest Manager. It is accessed by the CA for storing any incoming data, and by the SC for augmenting context-aware services.
U-Hopper System Architecture
Interest Manager (IM):
merges the user profile and the requirements originating from the hosted services, into interests , which are a description of the data requested by the user.
Content Acquisition (CA):
stores/update/removes data according to user preferences and service requirements.
Opportunistic Communication (OC) Unit
this engine transparently exchanges data among mobile nodes encountering on the move. Also it is in charge of reading nearby sensors. Such unit periodically searches available data sources, and takes care of all the necessary steps for gathering such information.
Information exchange SERVER CLIENT Retreive DATA Retreive DATA Store DATA Store DATA Open Connection Send INTERESTS Send DATA Send INTERESTS Send DATA Close Connection
U-Hopper in pills…
The middleware:
J2ME version + J2SE version
Leverage on Bluetooth for communication
Persistency: RMS on J2ME and MySQL Db on J2SE devices
Used as a pure middleware
Multiple applications with U-Hopper
P2P data exchange (ring-tones + advertisement)
Business cards exchange
Sensor data reading and exchange
Others (not yet implemented…)
U-Hopper application: mobile P2P Opportunistic Information Exchange BlueTooth Module Mobile Node MN Interests: U2 / Walk-On MN2 MN1 Interests: U2 / Walk-On
MSN: an analysis
Office environment test bed:
Interests collected through questionnaires
Contact pattern registered using U-hopper
21 participants for a 3 weeks period
MSN: The contacts pattern
Graph-based representation of the network of contacts
An edge exists between any 2 vertexes if contact time is at least 30 minutes per day.
MSN: emulation results
Nodes infection ratio has been evaluated:
Injecting packets in the network
According to users’ interests
On three different formats: text, music video
Packets have a varying TTL
MSN: introducing the social dimension
Questionnaires have been distributed:
To understand the real interests of users
To map networking interaction with real interests
Interests have been added to users profiles
A java simulator has been developed:
With real contact traces
Mapping them with real interests
We have defined a metric to understand impact of sociality into opportunistic networking
MSN: users affinity
Affinity :
Preferences are:
The resulting graphs
with affinity>0.75 =>
MSN: friendship based routing
Users exchange data if their interests match enough
Interests are weighted according to a predefined # of friends K that have those interests
K-nearest friends are selected for info diffusion
MSN: friendship routing results
We evaluated:
Network Infection Rate i.e. how much messages are propagated in the network?
Utility: upon receiving a message, how much does it much user’s interests?
MSN: the T9 service evolution
Opportunistic networking is used to have a service evolving in a distributed way
Users exchange services parameters
The service evolve in such a way that user perception of the service increases
We defined a mathematical framework to do that
The example used is T9 service where we showed dictionary evolution to better fit users’ request
MSN: the T9 service evolution
Fitness:
is defined as the satisfaction a user experiences when using a service
The higher the fitness the higher the degree of satisfaction
In the T9 example:
User fitness is low if he has to search to many time for a word
Fitness is high if all words come at first
Mobile opportunistic networking could help
If I meet a user with high fitness he sends me his dictionary
My dictionary enlarges and my fitness increases
We have defined an analytical framework and a simulation framework:
to understand how the T9 service could evolve with mobile interactions
The impact of mobility for increasing users’ fitness
Missing word: Bondone, falaise MSN: T9 distributed evolution TARGET TEXT : “Shall we meet tonight on the Bondone at 8.00 pm on the falaise ?” Opportunistic Information Exchange BlueTooth Module Mobile Node MN Bondone, falaise Has the words: Bondone,falaise Fitness is high MN5 Missing word: Bondone,falaise Fitness really low MN3
Conclusion
In this work, a system architecture:
has been defined, designed and implemented
to work in mobile pervasive environments
assuming no centralized connections
The designed system architecture:
solve the ITS scenario issues (routing and data management)
deals with the MSN challenges
Challenges of such scenarios have been faced and overcome from:
an analytical point of view
through an extensive simulation analysis
implementing a prototype and applying it to real world
0 comments
Post a comment