Pervasive Computing

1
Pervasive Computing

• Challenges, Capabilities and Responses
• February 26, 2002
• Martin Herman, Chief
• Information Access Division

2
Outline

• Pervasive Computing
• Challenges to NIST
• Program Goals
• Our unique position
• Current
• Projects
• Accomplishments
• Goals for the future

3
(No Transcript)
4
Pervasive ComputingWorking Definition

• Numerous, naturally accessible, integrated
  computing devices and sensors
• Frequently mobile (PDAs, smart phones)
• Often embedded (walls, furniture)
• Connected to ubiquitous and secure network
  infrastructure (wired and wireless)
• Omnipresent access to information and services
• Multi modal, distributed user interfaces
• That must enhance productivity

5
Challenges

• Pervasive Computing scope is vast
• Heterogeneous, distributed hardware and software
• Dynamic networking
• Large scale information access
• Distributed user interfaces
• Most companies focus on specific technologies
• Many immature, high risk, technologies
• Technologies rapidly growing - independently
• Lack of open standards, metrics, tools
  reference data
• confuse users and suppliers
• impede integration and interoperability
• difficult to measure quality and performance
• impede commitment of participants

6
Market Trends and Forces

• Smaller, cheaper, faster, computers sensors
• Enormous growth of Internet technology
  applications
• Growth of broadband wireless networking,
  nomadic devices
• Sensors increasingly networked, e.g microphones,
  cameras, process control, medical data management
• More embedded and mobile information appliances
• Perceptual interfaces being commercialized, e.g.
  speech, speaker, face recognition natural
  language processing

7
Pervasive Computing Environments
Distributed, Perceptual Interfaces (e.g., speech,
vision)
Pervasive Software (e.g., Exist simulation tool)
Integration Technologies (e.g., Smart Flow System)
Security (e.g., PKI crypto applications,
protocols, algorithms)
NIST Products, Outputs -gt Goals
Reference Data, Metrics, Software evaluation
simulation tools
Pervasive Computing Environments


Interoperability, conformance, performance tests
Test Beds
Pervasive Devices (e.g., smart cell phones,
wearable computers)
Information Access (e.g., multimedia retrieval)
Dynamic Service/Device Discovery (e.g., analysis
empirical measurement of service discovery
protocols)
Wireless Networking (e.g., analysis of
interference between Bluetooth WLAN devices)
8
Goals and Objectives

• Foster Pervasive Computing by providing industry,
  users, standards organizations, and academia the
  tools and tests to
• Identify and define product requirements
• Analyze and compare technical solutions
• Measure components and systems
• Develop high quality, correct, robust
  implementations
• Provide methodologies, standards, metrics, tests,
  and reference data needed by industry

9
Efforts in Industry and Academia

• Stanford U. iRoom
• MIT Intelligent Room Things That Think
  Oxygen Smart Room Wearable Computing
• Georgia Tech Aware Home
• CMU Pervasive invisible computing Automated
  surveillance Multimodal perceptual interfaces
• UC-Irvine, San Diego Cal-(IT)2 Wireless
  technology Mobile computing
• Rutgers Multimodal User Interaction Microphone
  Arrays
• University of Colorado The Adaptive House
• Many more

• Anderson Consulting - Active Environments
• HP - Sensor-based interfaces Data Appliances,
  Media Systems
• IBM - Pervasive Computing, Natural Interactivity
• Intel - OpenCV Home RF Working Group
• Microsoft - Intelligent environments, Easy Living
• Motorola - Human Interface Laboratory
• SUN - Java/Jini, Advanced Computer Interfaces
• Many more

but these efforts focus on component
technologies and specific systems, while NIST
effort focuses on interoperability, integration,
testing standards.
10
Conference Series 98, 99, 00, 01, 02
Pervasive Computing 2001
11
NISTs Unique, Neutral Position Allows us to
Develop

• Modular test beds for pervasive technologies
• Data transport, interoperability infrastructure
• Reference data production
• Densely populated wireless environments for
  protocol testing
• Functional performance evaluation
• Sensors
• Computing components (e.g., recognition modules)
• Wireless networking
• Pervasive devices
• User interfaces

• Common reference implementations for industry
• Sensor-based user interfaces
• Data architecture
• Service Discovery
• 2.4GHz compatibility
• Common metrics and test tools
• Metrics needed by companies
• Data annotation systems
• Impartial evaluations of systems, components, and
  protocols
• Conformance interoperability tests
• Simulation tools

Currently working on these tasks
12
CurrentFocus Areas
Pervasive Software
Pervasive Integration
Pervasive Networking
13
Pervasive Integration Goals

• Accelerate development deployment of smart
  space technologies
• Modular test bed for integration
  interoperability
• Sensors
• Devices
• Computing components
• Reference architecture to facilitate experiments
  interoperability in industry academia
• Develop
• Interoperability standards
• Tasks scenarios for testing
• Metrics for component and end-to-end technologies
• Collect and distribute test/reference data for
  multi-component systems

• Targeted smart space technologies
• Multi-modal user interfaces
• speech
• natural language
• visual input
• Multi-sensor integration fusion
• Integration with
• information retrieval technologies
• mobile devices and sensors
• network services
• Collaboration technologies
• Real time high throughput sensor data processing
• Collection of standard reference data

14
Reference Data from Sensor Arrays
Meeting Room Data
Single Molecule Metrology
15
Pervasive Integration Results

• Smart Flow test bed deployed to
• Rutgers CAIP Center
• MIT AI Laboratory
• Georgia Tech Aware Home
• Bell South
• SRI
• NIST Meeting Transcription project
• NIST Single Molecule Manipulation Measurement
  project for real-time sensor processing
• Integrated industrial components
• IBM speech recognition
• Intel OpenCV face recognition
• Wireless networking

• Unique sensor arrays for data acquisition
• Beam forming
• Source localization
• Acoustic/video sensor fusion
• Collection distribution of speech video data
  using Meeting Room
• Open source release of Smart Flow System
• Novel speech SNR metric to evaluate beam speech
  signal quality
• Smart Space research web site

16
Pervasive Software Goals

• Develop software models and tools to enable
  analysis, visualization, and measurements
• Model system architecture specifications
• Simulate pervasive environments

• Apply Architecture Description Language (ADL)
  models to service discovery protocols
• Evaluate ADLs as a tool for improving and
  extending software specifications
• Demonstrate utility of ADLs to analyze the
  robustness of distributed systems in response to
  dynamic change
• Provide models for analysis, comparison, and
  measurements of SDPs

• Develop EXiST Simulation Tool
• Object-oriented toolkit for creating and running
  simulations
• Provide ability to measure and test systems
  without having to build entire system
• Allow manipulation and measurement of different
  components as well as entire system
• Help identify bottlenecks and other performance
  problems

ADL computer-readable language for rigorously
defining systems architectures, i.e.,
components, their connections, and topology
17
Rapide Models Analyze Service Discovery
Protocol Dynamics
Topology
Scenario
Execute with Rapide
Topology
POSETs
Behavior Model
Consistency Conditions
Use metrics to Assess Correctness Performance
18
Pervasive Software Results

• EXiST Simulation Tool
• Completed feasibility and prototype of the tool
• Designed around XML-based web services (allows
  for portability, interoperability,
  extensibility),
• Initial performance measurements collected
  (consulting with Stat Eng Div)
• Documented concepts and designs in several
  papers/presentations

• ADL Service Discovery Model
• Conducted analysis of six service discovery
  protocols
• Explored correctness of Jini model against
  various consistency conditions
• inconsistency uncovered and presented to Sun
  Microsystems
• Developed generic models of 2- and 3-party SDP
  architectures
• Developed Metrics for comparison of SDPs in given
  environments
• Provided recommendations on improving next
  generation ADLs
• Documented results in several papers/presentations
  to SDP, Software Architecture, and DARPA
  communities

1 client, discovery
19
Pervasive Networking Goals

• Dynamic Service Discovery
• Analyze performance and scaling characteristics
  of Service Discovery Protocols (SDPs)
• Design methods and tools to measure scalability
  and performance of SDPs
• Design, simulate, and characterize self-adaptive
  mechanisms to improve performance of SDPs
• Analyze multiple service and device discovery
  protocols, e.g. Jini, UPnP, SLP, Salutation,
  HAVi, Bluetooth
• Implement and validate the most promising
  algorithms in publicly available reference
  software

• Coexistence of Wireless Devices operating in the
  2.4 GHz ISM band.
• Analyze and evaluate interference in the 2.4GHz
  ISM band caused by the coexistence of WPAN (
  Bluetooth, 802.15.x), WLAN (802.11.x), and sensor
  networks.
• Develop mechanisms at the MAC and PHY levels to
  reduce mutual interference
• Detailed MAC and PHY simulation models

20
Pervasive Networking Results

• Analyzed six discovery protocols and developed a
  unified model
• Established test bed system for Java/Jini and
  UPnP protocols, and took initial load and latency
  measurements
• Completed Rapide architectural-description
  language model of Jini and UPnP specifications
• Developed scenarios to assess correctness of Jini
  protocol design in the presence of node and link
  failures
• Created synthetic workload generation tools for
  emulating behavior of large scale, dynamic ad-hoc
  networking environments

• Modeled and evaluated mutual interference among
  Bluetooth and 802.11 networks
• - Using analysis, experimentation, and simulation
• - Metrics packet loss, access delay
• Techniques to compensate for interference
• - WLAN adaptive filters to suppress narrowband
  interference
• - Bluetooth MAC scheduling to avoid transmission
  in bad channels
• IEEE 802.15.2 Task Group on Coexistence
• - Developed recommended practice document
• - Editorship of specifications for 802.15.2

21
Performance Results
Bluetooth Performance
802.11 Performance
0 1 2 3
4 5 Distance between
Bluetooth Slave and 802.11 Mobile (meters)
Bluetooth MAC scheduling reduces the packet loss
to zero for both Bluetooth and WLAN systems
without increase in delay.
22
Future Plans and Opportunities

• Extend Simulation tools
• Integrated collection of tools and services
• Repository of measurements and data (from NIST
  and industry)
• Conformance test development
• Applicability of architecture description models
• Tech transfer Service Discovery Protocol models
  with tools
• Domains with dynamic reconfiguration (emergency
  and disaster response scenarios, homeland
  defense, DARPA)
• Enhance software and network testing program
  i.e real-time measurements for adaptive
  networking

23
Future Plans and Opportunities

• Extend coexistence mechanisms to support
  Bluetooth applications with Quality of Service
  requirements (voice video, MP3)
• Investigate coexistence of WLAN with additional
  WPAN technologies
• Low rate (802.15.4)
• High rate (802.15.3)
• Ultra wide band (802.15.3)

24
Future Plans and Opportunities

• Further deployment of Smart Flow system in
  different organizations and applications
• Collaborative communication and problem solving
  using smart spaces
• Evaluation methodologies
• User interface evaluation
• Emergency/disaster response scenarios
• Multi-modal test data collection
• Multi-modal, multi-sensor annotation methodology
• Evaluation methodologies
• Context-aware applications
• End-to-end systems
• Establish interoperability standards for
  integrated systems