Gaetano Borriello Department of CS

1
The Portolano/Workscape Expeditionin Invisible
Computing

• Gaetano BorrielloDepartment of CSEUniversity
  of Washington
• DARPA VisitArlington, VA29 November 1999

www.cs.washington.edu/research/portolano
2
Invisible computing

• Mobile, ubiquitous, pervasive, but also . . .
• fast, if not effortless, learning curve
• tailored to the task at hand (hw sw)
• negligible maintenance overhead
• robust and reliable
• safe and trustworthy
• Low cognitive load on the user
• dont distract when you dont have to
• users attention is the valuable resource

3
Expedition goals

• Invisible user interfaces
• computing, not computers
• user-centered and task focused
• devices of all shapes, sizes, and functions
• Universal connectivity
• diverse physical layers
• middle-ware and app. dev. to support distr.
  systems
• power thrifty, intermittent connections
• Intelligent services
• instrument physical environment, people, and
  objects
• automatic agents to make them all communicate
• off-load as many as possible of users
  tasks/concerns

4
DARPA missions

• Highly relevant to DoD needs for the 21st century
• robust communication systems
• multi-modal sensor arrays
• automatic surveillance agents
• command and control
• mission planning
• personnel training
• robust and more diverse information/service space
• Many consumer applications
• leverage rapidly evolving consumer electronics
  base
• opportunities to really build, deploy, test, use
  our creations
• latch on to consumer cost curves for
  infrastructure

5
Expedition personnel and their expertise

• Larry Arnstein embedded systems, applications
• Gaetano Borriello embedded systems, tools
• Mari Ostendorf voice recognition and synthesis
• Roy Want embedded systems, devices, MEMs
• Tom Anderson networking
• Alon Levy databases and network-based services
• Karin Petersen storage systems/services, UIs
• Dan Russell human-computer interaction, AI
• David Wetherall active networks and protocols

6
A team effort
studentsinterns

• UW CSE Anderson, Arnstein, Borriello,
  Levy, Ostendorf, Wetherall
• Xerox PARC Want, Russell, Petersen
• Support
• DARPA-ITO Expeditions
• NSF-CISE/CCR embedded sw
• NSF-CISE/CRCD edu modules
• Intel Corporation
• Ford Motor Company
• Other DARPA interest
• Active Nets/Names - Maughan/Maeda (ITO)
• Labscape (UW Cell Systems) - Skurnick (DSO)

• Collaborators
• Intel Research Labs
• Microsoft Research
• HP Labs
• IBM TJ Watson
• ICSI

7
Our strengths and foci

• Devices
• deployment experience
• user interfaces
• fashioning into invisibility
• Middle-ware
• networking (protocols to active nets/naming)
• data services
• application development/deployment/monitoring
• Applications
• real users in useful activities
• multiple domains
• strong collaborators in each

8
Agenda

• Introduction - Borriello
• Connectivity/Middleware/Data Services -
  Borriello
• Devices/Infrastructure - Want
• One of our Driver Applications (Labscape) -
  Arnstein
• Speech Interfaces and Services - Ostendorf
• Expedition organization into options - Borriello

9
A Visionary

• Prince Henry the Navigator (1394-1460)
• Patron of explorers, one of the earliest
  geographers
• Established school for navigatorsin 1450 at
  Sagres

10
The Dawn of the Age of Exploration

• Portolano charts - first truerepresentations of
  coastlines
• Published in time for next voyagefurther down
  African coast
• Repository at Sagres school

11
Workscapes

• Capture, browsing, knowledge extraction
• Fluid style that doesnt require user attention

displays
portals
sensors
interactors
12
Portolano/Workscape charting the new territory
ofinvisible computing for knowledge work

• New Ideas
• Invisible user interfaces
• user movement, proximity, location, etc.
• data fusion to extract user intent
• Data-centric networking
• data code to get data where it needs to go
• ultra-low-power and privacy lead to new
  tradeoffs
• Distributed services
• layered services on distributed database
• agents to filter actuation requests and link
  layers
• Application deployment/delivery
• new programming environment for distr. code
• synthesis approach to comm./coord. code

agents
sensors
actuators
Distr DB
Gaetano Borriello, Tom Anderson, David
Wetherall University of Washington Roy Want,
Daniel RussellXerox Palo Alto Research Center
Schedule

• Impact
• Get user interfaces out of users way
• low cognitive load on user
• form-factors tailored to task-at-hand
• Increase data security and robustness
• replication throughout network
• redundant data sources and agents
• Simplify composition of network services
• data-centric service discovery
• complexity through layering
• Make it easy to create new applications
• automatic synthesis of most complex code
• system-level rather than component-level view

Develop data-centric networking protocols built
on activenetworking infrastructure
Begin analysis of user experience
Choose severaldemonstrationdrivers
Second iteration of deploymentand evaluation
Sep 99 Start
Sep 00
Sep 01
Sep 02
Develop initialUI elements
Begin deploymentof devices in context
Deploy distributed DB for horizontally-layered
services
Release infrastructuresoftware to wider community
13
Invisible computing photography scenario
14
Expedition assumptions

• Expanding consumer market for task-specific
  devices
• World-wide information/computation utility to
  provide basis for communicating/computing
• Moores law doesnt apply (and probably never
  will)
• power sources
• user mind-share
• Moores law applies (at least for a while
  longer)
• bandwidth
• computation cycles
• storage

15
Influencing trends

• Non-productive interaction with computers leads
  to a yearning for
• ease of use, universal plug-and-play
• trustworthiness
• ubiquity (spatial and temporal)
• Communication standards lead to interoperability
• wireless Bluetooth, IrDA, PAN
• wired TCP/IP, active nets
• discovery services
• Technology leads to new devices
• embedded sensors/actuators
• low-power and power harvesting
• task-specific form factors

16
Expedition challenges
devices

• Intentional user interfaces
• Low-power architectures/devices/communication
• Application partitioning for low-power
• Application development/deployment
• Data-centric networking
• Scalable discovery/directory services
• Composable, horizontally-layered services
• Distributed databases and data integration
• Data mining and database layering

middle-ware
services
17
Expedition planning methodology
User Services
Data Services
Middleware
Connectivity
Devices
18
Application domains (and collaborators)

• Labscape (UW Cell Systems Initiative)
• instrumentation of the workplace
• collect data to replay meetings/experiments
• data mining to support investigations/recollection
  s
• Personal devices/networks (Microsoft Research)
• body-area networking (RF and skin)
• borrowable, scrap devices
• Here-to-there (Intel and Ford)
• in-building/in-room location tracking
• continuous access to data
• management of personal devices and services

19
Research matrix
Here-to-there Body-area/ Borrowable Labsca
pe
Distributed Systems Human-Computer
Interaction Speech Technologies Operating
Systems Data-Centric Networks Devices Realization/
Infrastructure Knowledge Capture/Representation Da
tabases
20
Agenda

• Introduction - Borriello
• Connectivity/Middleware/Data Services -
  Borriello
• Devices/Infrastructure - Want
• One of our Driver Applications (Labscape) -
  Arnstein
• Speech Interfaces and Services - Ostendorf
• Expedition organization into options - Borriello

21
Networking many small mobile devices

• Moores law does not apply to batteries
• Connections will always be limited in range and
  speed
• Intermittent connections will be the norm
• Data will have to hop from device to device
• TCP/IP isnt appropriate for everything

22
Data-centric networking

• Data has a life of its own, carries its own code
• Uses computational resources of nodes at boundary
  between wireless and wired worlds

wirelesswired
web (data storesand services)
devicegateway
bodyserver
23
Connection-centric networking
24
Data-centric networking
25
Data-centric networking

• Supports low-power devices
• drop-in and run saves transceiver power
• ambient power harvesting becomes possible
• fosters power-oriented application partitioning
• 3-point acknowledgements
• receipts match-up with acks on return to base
• enhanced privacy/security (also through
  anonymizers)
• Requires active networking infrastructure
• ship code in data bundle
• need business model for charging for resources

26
Networking infrastructure

• Active names flexible name resolution system
• a new DNS with names unbound to physical
  addresses
• supports new Internet services, e.g., server load
  balancing
• operates above ANTS
• composable location-independent model
• Jini discovery services
• scalability issues in multi-cast - meta-services
• small/ultra-low-power/legacy devices need proxy
  services
• ANTS widely used active networks infrastructure
• new deployment strategies to support innovation
  in services
• active data trust and security issues
• code migration to most appropriate network nodes

27
Some middle-ware research problems
Components
Self-describing active data

• Development
• Deployment
• Monitoring
• Updating
• Distributed data-bases

Active networks
Full-system perspective
28
Model-based Design

• Pluggable reusable components
• Data-flow and control-flow plugs
• events and state changes

objects state
29
Development - partitioning/synthesis
synthesize communication/coordination code
30
Deployment - plug-and-play at all levels

• Devices search out and find their own code
• Data finds its own way through the network
• self-describing, includes code for
  replication/discovery/transformation/fusion
• active networks and discovery services
• intermittent connections
• data-specific policies for replication/forwarding
• Architecture of cheap sensors, simple network
  nodes, computational power in back-end services

31
Monitoring - how do you know its working?

• Functional code is not enough
• Checking code is needed
• Determine failure points
• find alternatives
• change data policies
• retransmit from redundant stores
• inform user in a meaningful way

32
Distributed (layered) data services
agents(pattern detectors)
sensors(input, e.g., location, web info, medical)
distributeddatabase
actuators(output, e.g., alerts,web buys, on/off)
33
Agenda

• Introduction - Borriello
• Connectivity/Middleware/Data Services -
  Borriello
• Devices/Infrastructure - Want
• One of our Driver Applications (Labscape) -
  Arnstein
• Speech Interfaces and Services - Ostendorf
• Expedition organization into options - Borriello

34
What is a Workscape?

• Real workplaces go beyond the computer
• desks walls filing cabinets
• offices and Laboratories
• halls and conference rooms
• buildings, cars and in the field
• A workscape is all aspects of a working
  environment, both physical and virtual

35
The Physical World is Complex
lab mtg
B Y
B R
use
Bs office
B Y
Y G R
Ys office
36
Application 1 Commandscape

• Spontaneous formation of a command post
• Temp Military Base (take over an old building)
• Xerox goes to COMDEX (booth, press room, keynote
  stage, hotel rooms)
• Some issues
• Collection of information
• Distribution of commands
• Making sense of large amounts of data

37
Application 2 Labscape

• Lab results are hard to reproduce
• need to track everything
• May need to collaborate with remote groups
• Tacit collection and distribution
• Data representation and presentation is key to
  making use of complex data-sets

38
Application 3 Protoscape

• Design and prototyping environment
• Groups could be scattered cross-country
• e.g. ID Bleck Design group in Boston working with
  Xerox PARC
• state of the hardware prototypes (milling,
  painting)?
• CAD tool status
• transfer of proofs
• shipping information status
• billing for materials and services

39
Workscape Infrastructure

• Sensing
• Hard sensors field sensor data (physical)
• Soft sensors computational sensor (virtual)
• Architecture optimizing data collection, data
  fusion and distribution from/to various sources
• Sense making
• Presentation

40
Previous work Ubiquitous Computing

• Ubiquitous Computing Phase I 1988-1995
• Parctab (inch) , ParcPad (foot), Liveboard (yard)
• DARPA contract DABT63-91-C-0027

Wearable
Carryable
Fixed
41
Previous Work Parctab Project
Diffuse-IR Wireless micro-cellular PDA
42
Previous WorkIn-Building Location Systems
Active Badge
Olivetti Research Hardware Xerox PARC/EuroPARC
AIR project Ubicomp
43
Novel Workscape Components

• Electronic tags with sensing
• Ad hoc networks
• dynamic configuration and location-awareness
• active data
• Programmable hardware devices
• Reconfigurable systems
• just-in-time hardware solutions
• network programmable FPGAs

44
Augmenting the Computer
Pilot pressure interface
Cassiopeia with pressure strip
Pilot with 1D tilt interface
Xerox PARC/FX Hikari PDAwith 2D tilt-interface
45
Tagging and Sensing

• Most electronic tags just have a unique ID
• Modern RFID tags contain memory (256 bits)
• A tag can be more like a network node and also
  sense its environment

E-TAG IDENTITY (unique) STATE (static and
active) COMMUNICATION
46
Augmenting the World with invisible interfaces
47
Electronic tags novel approaches

• Inductive RFID (traditional approach)
• Capactively coupled RFID
• Infrared beacons
• Infrared transponders
• Location based on signal strength
• Location based on signal timing

48
Inductive RFID

• Inductively powered by reader
• one Coil
• one Chip
• encapsulation

49
Capacitively coupled RFID

• Motorola Bistatix Tags
• Dipole or Monopole coupling
• Orientation tolerant
• Does not need a resonant circuit with high Q for
  maximum coupling
• Antenna can be high resistance (gtgt100R/sq.)
• Demo
• pencil lead graphite antenna
• Tear-off pieces of antenna and it still works

50
Motorola Bistatix
Graphite antenna
RFID Chip
51
Ad hoc Networks

• Sensors/node/tag deployment for ad hoc nets
• fixed position
• mobile hosts (equipment, documents, people)
• scattered
• Tags/sensors become network nodes that can be
  temporarily disconnected
• How do we deal with distributed reconciliation
  for sensed data that is old
• Sensor fusion in ad hoc network

52
Spot-on RF location
Ethernet
RFbasestation
Spot-on mobilenode
Accelerometer RF signal strength
53
Whiteboard Application
An extended Diagrammatic User Interface (DUI)
used to connect components of a conference room

Whiteboard
Monitor
Meeting Notes for today
Video Tape
Speakers
Volume
54
Infrared Tags

• CCD cameras are sensitive to IR

55
In-band IR signaling
IR FLASH
Frame 1 , 2, 3, 4 ..
Phicon 1
1
1
1
0
Phicon 2
1
0
1
1

• Spatial coding for each phicon (x, y)
• Bits are transferred frame by frame (time coded
  in image)
• Advantage lots of phicons supported
• Disadvantage slow data rate 15bps

56
Out-of-band IR signaling
FLASH !
Y AXIS
RS232
X AXIS

• Flash at (x,y) in camera is correlated with
  signal on RS232
• Signal ID is pulse positioned coded
• Advantage Fast data rate for each phicon
• Disadvantage small number of phicons supported

57
Keychain Computing
58
Micro-minder technology

• Phicon possibilities

• Minder Device

Programmable in C Download via IR from PC
59
Network Programmable Hardware
FPGA
Physical Interface (RF, cable-dock)
Micro-webserver
Ethernet
Mobile Devices
Mobile Devices
Mobile Devices
60
Programmable Hardware

• Examples of flexible behavior (function
  programmable)
• compression
• filter
• mediate
• track
• analyze
• gating sensor data at the sensor (biological
  analogue)

61
Sensor Fusion at all levels

• Low-level data fusion in infrastructure
• High-level data fusion from databases
• Different views of data for many applications
• local or global sharing/collaboration
• Reinterpretation in different contexts
• Data from Hard sensors
• Data from Soft sensors

62
Knowledge-creating Environment
63
Knowledge-sharing mediums
64
Summary

• Portolano / Workscapes will be ideally placed to
  take advantage of state-of-the-art technologies
  that enhance work practice, through ...
• embedded computation
• automatic identity
• location derivation
• just-in-time hardware
• wireless connectivity

65
Agenda

• Introduction - Borriello
• Connectivity/Middleware/Data Services -
  Borriello
• Devices/Infrastructure - Want
• One of our Driver Applications (Labscape) -
  Arnstein
• Speech Interfaces and Services - Ostendorf
• Expedition organization into options - Borriello

66
LabScape - one of our driver applications

• Biology is a hard science with a soft
  infrastructure
• capture and use of knowledge is key
• from loosely connected to highly integrated
  collaboration
• invisible infrastructure for building knowledge
  base

67
LabScape is an Ideal Application Driver for
Portolano

• Sophisticated butnon-IT user base
• Failure of LIMS desktop applicationsapproach
  to laboratoryautomation
• pen and paper labnotebooks still the norm
• Rich but rigorous ontology
• Green-field opportunityto explore invisible
  computingin the Cell Systems Initiative
  facility

68
LabScape Major Components

• Descriptive Model The Knowledge Base
• heterogeneous
• globally linked reviewed and raw
• semantically precise, without excluding narrative
• contradictory and negative information
• Experiment Manager Knowledge Capture and Use
• remote utilization of capital intensive resources
• capture data not known to be important
• collaboration across scientific community
• build the database by doing the work!
• access and augment the knowledge base

69
Scenario The Scientific Process

• Tom and Aparna are collaborative molecular
  biologists
• Aparna gets a hit on her standing query someone
  appears to have validated a hypothesis about the
  synthesis of a particular protein using a new
  experimental technique. Its Tom.
• Aparna wants to reproduce the experiment. Gets
  Toms protocol and assigns a student to perform
  the experiment.
• The student sees something questionable in Toms
  protocol and changes the experiment on the fly.
• Aparna reviews her students lab techniques and is
  pleased. Adds new hypothesis and publishes the
  raw data.
• Tom notices a new reference to his experiment and
  learns that Aparna claims to have invalidated his
  results. A quick query shows exactly what was
  different between the two protocols. Tom is
  pleased that his experimental techniques are
  being repeated.
• Tom sends a message to Aparna, thanks her the
  improvement and feedback. Tom reduces the
  confidence level of his hypothesis but does not
  remove it from the system. They publish their
  results by submitting the knowledge base for
  review.

70
Scenario
71
Scenario now cast in invisible terms
Invisible Computing Infrastructure
72
Portolano Research Issues

• User Interface
• scientific process, domain knowledge
• multi-mode
• location-tracking
• speech I/O
• Infrastructure
• raw data capture
• tagging and tracking
• fusion
• equipment control
• data storage/manipulation
• Distributed Services
• experiment re-creation
• data mining/query
• master/apprentice
• maintenance
• contextual whiteboards

Experiment Manager
Descriptive Model
73
University of WashingtonCell Systems Initiative
School of Medicine
College of Engineering
Cell Systems Initiative
Isis
Merck
Sun
Immunex
DARPA
SAIC
Aurora
74
Agenda

• Introduction - Borriello
• Connectivity/Middleware/Data Services -
  Borriello
• Devices/Infrastructure - Want
• One of our Driver Applications (Labscape) -
  Arnstein
• Speech Interfaces and Services - Ostendorf
• Expedition organization into options - Borriello

75
Speech as the Interface . and
the Information

• Mari Ostendorf -- UW EE dept
• Nelson Morgan -- ICSI

76
Speech and Invisible Computing

• As part of the computer interface
• natural command and control
• universal information access(Speech interfaces
  are invisible by being natural.)
• As an information source
• broadcast news browsing
• instrumented lab or meeting room (dedicated
  speech server and mic array)
• portable meeting transcriber (speech-corder)

77
Labscape control, capture and replay

• Instrument control
• Speech recording provides
• documentation of motivation for each step
• alternative record for aspects that are difficult
  to instrument
• Random access of hours of speech requires new
  indexing technology
• Experiment replay for training requires speech
  synthesis in style of an expert

78
Speech Corder No More Note Taking

• Record conference call or meeting
• Annotation for indexing words,topics, speakers,
  QA links, linksto notes and presentation
  materials
• Speech queries for random access
• Play me Janes answer to Joes questionabout
  the IP agreement.
• What graph was showing when Jeffquestioned the
  results?
• Who spoke in favor of his proposal?
• Review action items.

good point!
action item
dont agree, check data
79
Communicator Human-Computer Interaction

• Information access by phone (weather, flight
  info, news/mail/meeting browsing)
• Interactive dialog for collaborative
  human-computer problem solving (e.g. travel
  planning)
• Our new twist dialog can be multi-party
• Speech needed for both input and output, for
  seamless switching between speech and text

80
Core Technologies Needed

• Robust speech recognition and understanding
• Natural speech generation
• Dialog tracking
• Speech indexing for retrieval
• Information extraction from errorful data
  expedition effort related effortsadditional
  funding from NSF, German SmartKom,
  collaborations with MITRE, SRI, MIT, ...

81
Speech recognition is NOT a solved problem!

• 1999 large vocabulary recognition results
• Read versus spontaneous speech
• yeah yeah I've noticed that that thats one of
  the first things I do when I go home is I either
  turn on the t v or the radio its really weird
• Head-mounted vs. desktop mic 10 70 WER
  for a speaker-trained commercial system

82
ASR Research Challenges

• Human-directed speech
• Noisy, reverberant environments
• Coding for recognition in wireless environments
  (robust to packet loss)Reminder speech
  recognition is a critical link in human-computer
  dialog -- if the inputs broken, then nothing
  else matters.

83
Robust Speech Recognition - ICSI

• Robust feature processing (filtered features) and
  multi-stream recognition
• Recent results in ETSI distributed ASR standards
  competition

84
Robust Speech Recognition - UW
CMI/DMI plot of spectral features
Extended Markov Model
85
Coding for Wireless Communications and
Distributed Recognition Agents

• For applications where speech is coded for
  wireless transmission .
• Design the code
• to optimize recognition performance
• to have graceful degradation with packet loss
  using forward error correction
• Design the recognizer to use code directly for
  increased efficiency

86
Speech Generation

• Speech synthesis is not a solved problem!
  Synthetic Natural but is critical for
  user acceptance
• Approach improved prosody prediction for HCI
• domain-dependent, automatically trained
• using linguistic annotation from a text generator
  and dialog manager
• Integration of prosody prediction with unit
  selection for concatenative synthesis

87
Dialog Tracking

• Goal recognize dialog acts such as info request,
  help request, error correction, .
• Motivation
• More accurate speech recognition (dynamic
  language models), more appropriate responses
• Key problems
• Identifying topic changes, error correction
  sub-dialogs
• Resolving intention ambiguities (e.g. OK, thanks)
  using prosodic cues

88
Speech Indexing

• Goal navigating meetings, demonstrations etc.
  in query-based access (not simply IR on ASR
  outputs!)
• Annotation for indexing using
• automatic transcription with confidence
  annotation
• speaker and topic tracking
• dialog act labels (e.g. question, answer,
  clarify)
• reference linking within and across topics
• relative timing, possibly spatial orientation

89
Discourse Tracking and Indexing -- Whats New?

• Explicit modeling of recognition errors
• Integrating acoustic (prosodic) cues with wording
  cues
• Explicit use of timing
• Use of statistical language models in IE to aid
  reference and topic linking
• Automatic analysis for multi-party case

90
Summary

• Speech processing technology provides an
  interface and an information source
• Applications
• Labscape
• meeting browser
• travel task
• UW/ICSI core speech technology research
• robust speech recognition, distributed (wireless)
  processing
• dialog, multi-party discourse tracking and
  indexing

91
Agenda

• Introduction - Borriello
• Connectivity/Middleware/Data Services -
  Borriello
• Devices/Infrastructure - Want
• One of our Driver Applications (Labscape) -
  Arnstein
• Speech Interfaces and Services - Ostendorf
• Expedition organization into options - Borriello

92
Portolano/Workscape - Summary

• Major goals
• keep information informal, graceful and fluid in
  use
• content easily captured and never lost (unless
  desired)
• applications with rich set of inputs/outputs
• drive down cost of creation, access and use
• leverage technology curves
• Major opportunities
• create environment of space, services, portals
• data-centric protocols for intermittent and
  mobile use
• distributed databases
• high-level distributed application development

93
Prototype products in invisible computing

• Preferences holder with short-range RF
  communication
• GPS-enabled notes on a PDA
• Voice-programmable VCR
• Shoe-embedded pedometer to calculate distance
  walked/run
• Real-time rapid transit information on a PDA
• Adaptive remote control for home appliances
• Novel input modes for PDAs (tilting, pressure,
  orientation)
• PDA Calendar that checks traffic conditions and
  plans best route
• Plug-and-play home automation IR
  location-sensing RF to server
• Training bike with GPS, speed, heart-rate
  sensing, RF to PDA
• Medical tricorder logs blood-pressure/heart-rate/b
  lood-ox
• Prescription entry and mgmt system (PDA, service,
  pager)

94
Possible exploratory projects

• Body-area server and protocols
• Borrowable PDAs (scrap devices)
• In-building location tracking
• Virtual neighbor modular home sensor
  architecture
• Software partitioning/mapping app. dev. and
  depl.
• Embedded web servers/gateways
• Generalized proxy services
• Active networks enhancements
• Labscape infrastructure
• Physical icons, tags, and sensors
• Integrating speech with sensors/tags
• Meeting browser
• Speaker/language modeling/identification

95
Base program

• User Interfaces
• sensors, intentional UIs, multi-modal, invisible
• Network Infrastructure
• data-centric networking, active data/networks
• Distributed Services
• distributed data stores, horizontal service
  organization
• Support initial investigations/feasibility
  studies
• Lay the infrastructure on which to build
  experiments
• Limited test-beds/applications

96
Program options

• Portolano infrastructure and its distribution
• User and usage studies of wide deployment
• Ubiquitous large displays
• Application development environment
• Creation of a set of workscape application/device
  suites
• Modular sensor system
• Location sensor technologies
• Novel automatic identification technologies
• Flexible active network infrastructures
• Applied demonstrations of sensor fusion
• Extended information-space management
  infrastructure
• A new user experience for distributed information
  flows
• Making sense of complex information from
  distributed information sources

97
Conclusion - research impact

• A big opportunity/challenge in information
  technology
• component-based design (devices to services)
• layered services and databases (horizontal, not
  vertical)
• self-describing data and active networks
• self-monitoring, self-updating systems
• Synergy of many disciplines
• embedded systems/sensors
• software engineering
• active networks
• low-power design
• speech technologies

• operating systems
• databases
• context-aware computing
• user interfaces
• development environments

98
Conclusion - DARPA missions

• Many consumer applications
• leverage devices and infrastructure
• opportunities to make real and put to use
• Highly relevant to DoD needs for the 21st century
• robust communication systems
• multi-modal sensor arrays
• automatic surveillance agents
• command and control
• mission planning
• personnel training
• robust and more diverse information/service space

99
Summary

• Creating a new world of invisible computing
• where computing fades into the background
  (Weiser91)
• low cognitive load
• Integrated approach from applications to I/O
  devices with network infrastructure and databases
  in between - strong track record
• Three application domains to drive vertically
  integrated research - strong collaborators/users
• www.cs.washington.edu/research/portolano