31
普普普普普普 Ubiquitous System Software Managing Uncertainty:Model ing Users in Location-Track ing Applications Present 普普 普普普 :一 Chang Yung-Chang

普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Embed Size (px)

Citation preview

Page 1: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

普及運算報告Ubiquitous System Software

Managing Uncertainty:Modeling Users in Location-Tracking Applicati

onsPresent :研一 張永昌

Chang Yung-Chang

Page 2: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Outline

Ubiquitous System SoftwareManaging Uncertainty:Modeling Users in L

ocation-Tracking Applications

Page 3: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

普及運算報告Ubiquitous System Software

Present :研一 張永昌Chang Yung-Chang

Page 4: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Outline

Introduction THE MOST SENSED CAMPUS MONITORING EARTHQUAKEINDUCED

LOADING WITH CAMERA NETWORKS MIN: MIDDLEWARE FOR NETWORK-

CENTRIC UBIQUITOUS SYSTEMS DESIGNING UBIQUITOUS SYSTEMS

THROUGH ARCHITECTURAL REFLECTION INTERACTION METAPHORS APPLICATION MODELING FOR CONTEXT

AWARENESS

Page 5: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Introduction

THE MOST SENSED CAMPUS MONITORING EARTHQUAKEINDUCED

LOADING WITH CAMERA NETWORKS MIN: MIDDLEWARE FOR NETWORK-

CENTRIC UBIQUITOUS SYSTEMS DESIGNING UBIQUITOUS SYSTEMS

THROUGH ARCHITECTURAL REFLECTION INTERACTION METAPHORS APPLICATION MODELING FOR CONTEXT

AWARENESS

Page 6: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

THE MOST SENSED CAMPUS

THE MOST SENSED CAMPUSMichael W. Bigrigg and H. Scott Matthe

ws, Carnegie Mellon Universitymost wired→most wireless→most sensed

Page 7: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

MONITORING EARTHQUAKEINDUCED LOADING WITH CAMERA NETWORKS

MONITORING EARTHQUAKEINDUCED LOADING WITH CAMERA NETWORKS

Tara C. Hutchinson and Falko Kuester, University of California, Irvine

Page 8: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

MONITORING EARTHQUAKEINDUCED LOADING WITH CAMERA NETWORKS

The investigation has two primary objectives : Characterize the seismic response of an import

ant class of equipment and building contents Study the applicability of tracking this response

using arrays of image-based monitoring systems

Page 9: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

MONITORING EARTHQUAKEINDUCED LOADING WITH CAMERA NETWORKS

Exploits several issues in designing networked sensing systems for field applications: Viability of high-speed networks of sensors under

adverse conditions (in this case, earthquake loads) Communication with a variety of different sensor types Interpretation capacity of the sensed information (by a

remote user) Network latency and failure tolerance under high-

demand conditions (high rates of acquisition, through adverse conditions)

Page 10: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

MIN: MIDDLEWARE FOR NETWORK-CENTRIC UBIQUITOUS SYSTEMS

Lu Yan, Turku Centre for Computer Science and Åbo Akademi University

MIN=Formal Methods in Peer-to-Peer Network

Page 11: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

MIN: MIDDLEWARE FOR NETWORK-CENTRIC UBIQUITOUS SYSTEMS

systems require A self-organizing infrastructure Dynamic topology A hop connection Decentralized service Integrated routing Context awareness

Page 12: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

DESIGNING UBIQUITOUS SYSTEMS THROUGH ARCHITECTURAL

REFLECTIONDESIGNING UBIQUITOUS SYSTEMS TH

ROUGH ARCHITECTURAL REFLECTION

Francesca Arcelli, Claudia Raibulet, Francesco Tisato, and Marzia Adorni, Università degli Studi di Milano-Bicocca

Page 13: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

DESIGNING UBIQUITOUS SYSTEMS THROUGH ARCHITECTURAL

REFLECTION Several relevant features :

complex multimedia multichannel mobile distributed systems

features : context awareness Location awareness self adaptation service orientation quality-of-service support awareness negotiation capability(to solve conflict resolution)

Page 14: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

DESIGNING UBIQUITOUS SYSTEMS THROUGH ARCHITECTURAL

REFLECTIONWe’ve designed a reflective architecture fo

r multichannel adaptive information systems (the MAIS project).

Page 15: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

INTERACTION METAPHORS

INTERACTION METAPHORSChristoph Endres, German Research Ce

nter for Artificial Intelligence Andreas Butz, Munich University, Germany

Page 16: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

INTERACTION METAPHORS

The FLUIDUM project =Flexible User Interfaces for Distributed Ubiquitous Machinery

WIMP=Windows 、 Icon 、 Menus 、 Pointing devices

FLUIDUM addresses instrumented environments at three different scales—the desk, room, and building levels 。

Page 17: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

APPLICATION MODELING FOR CONTEXT AWARENESS

APPLICATION MODELING FOR CONTEXT AWARENESS

Maja Vukovic and Peter Robinson, University of Cambridge

Page 18: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

普及運算報告 Managing Uncertainty:Modeling User

s in Location-Tracking Applications

Present :研一 張永昌Chang Yung-Chang

Page 19: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Outline

IntroductionModeling usersCollecting user dataBuilding the user modelUsing Bayesian networksPerformance issuesExperimental resultsDiscussion

Page 20: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Introduction

Applications : Track elderly people Provide targeted advertising to mobile users track moving objects

Page 21: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Modeling users

main variable types : Temporal variables represent when events occur, includ

ing the time of year, day of the week, and time of day. Spatial variables represent possible RU locations, such

as a building, town, certain part of town, certain road or highway, and so forth.

Environmental variables represent things such as weather conditions, road conditions, and special events.

Behavioral variables represent things such as typical speeds, resting patterns, preferred work areas, and common reactions in certain situations.

Page 22: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Collecting user data

Divide the data into two categories : User-specific data consists of personal informati

on or trip-related information Environment-specific data describes the differe

nt artifacts of the environments ( weather conditions 、 traffic conditions 、 special events taking place )

Page 23: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Building the user model

Common ways to build user models : Machine learning Predicate logic First-order logic

Page 24: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Building the user model

Page 25: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Using Bayesian networks

Values : Event Time of day Source Destination Weather conditions Route Speed

Page 26: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Using Bayesian networks

Page 27: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Performance issues

We would maintain the BN, which involves updating the probabilities associated with each node based on new observations, and we’d perform inference given some observation.

Page 28: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Experimental results

simulation : used these typical speeds to create the user spe

eds during a trip as follows: 45 percent of the time, the speed should be within 10 of

the RU’s typical speed under the current circumstances. 25 percent of the time, the speed should be within 20 of

the RU’s typical speed under the current circumstances. 20 percent of the time, the speed should be within 50 of

the RU’s typical speed under the current circumstances. 10 percent of the time, the speed should be within 100 o

f the RU’s typical speed under the current circumstances.

Page 29: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Experimental results

Experimental design 100,000 trips and performed 10 queries on eac

h trip. the experiment for RIs 5, 10, 15, 20, 30, 50, 100,

150, and 200 time units.

Page 30: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Experimental results

The figure shows that LSR performed better when the RI was less than 12 time units, at which point the two techniques performed equally well.

We can see how our approach outperforms LSR, especially with high RIs. Our technique at RI 100 and 200 performs better than LSR at RI 50 and 100, respectively.

Page 31: 普及運算報告 Ubiquitous System Software Managing Uncertainty:Modeling Users in Location-Tracking Applications Present :研一 張永昌 Chang Yung-Chang

Experimental results