Smart Sensors Intelligent Systems by psf35982

VIEWS: 28 PAGES: 41

									Smart Sensors & Intelligent Systems
“Design at the Edge of Chaos”




Technology Trends Seminar
August 9 2004




Dr Tony Farmer
CSIRO Industrial Physics Division & ICT Centre

                                          www.csiro.au
                                         www.csiro.au
Outline



  Introduction
  Concepts
  Multi-Agent Approach
      Distributed Intelligence
      Emergent Behaviour
  Demonstration Spaces
      Structural Health Management – AAV
      Distributed Energy Management & Control
      Self-Assembly
  Summary




                                                www.csiro.au
 Introduction

We are interested in:

     The design and performance of large networks.
     These may be sensor networks, comms networks,
     processing networks, ….
        small, fast, cheap, embedded devices
        pervasive, ubiquitous, wireless
        large, long-lived systems
        useable by non-experts
        possibly resource-constrained
        autonomous
     They may be fixed, mobile, or both and may be on scales
     ranging from nm to km.




                                                     www.csiro.au
Concepts



  Networks of the future will be large, complex systems.

  They will need to be scalable, robust, fault tolerant,
  secure and stable.

  They will need to make decisions - based on sensed data,
  inferred information, system knowledge and learnt knowledge


  Application areas are likely to require dynamic,
  reconfigurable, ad hoc networks (viral networks)
  with the ability to self-assemble.




                                                           www.csiro.au
Concepts



  They are likely to be complex systems and exhibit
  emergent behaviour

  Incremental advances in technology are unlikely to
  provide appropriate solutions as network size
  increases.

 A disruptive technology is required,

  however, near-term steps along the way will, of
  neccessity, involve conventional approaches.




                                                    www.csiro.au
Resources & Links



    Intelligent Systems Research Group

10 staff / 6.2 EFT in              12 staff / 8.9 EFT in
         CIP                           ICT Centre


    Robotics Group - Brisbane

                    16 staff / 15.5 EFT
                      in ICT Centre

    Others

CMIT, CMIS, CLW, CLI, CTFT, CEM                            NASA
NICTA, UNSW, USyd, UQ, MacqU, UTS                          Boeing


                                                           www.csiro.au
 Resources & Links

Staff

        Intelligent Systems Research Group

 10 staff / 6.2 EFT in              12 staff / 8.9 EFT in
          CIP                           ICT Centre


        Robotics Group - Brisbane

                     16 staff / 15.5 EFT
                       in ICT Centre

        Others

 CMIT, CMIS, CLW, CLI, CTFT, CEM                            NASA
 NICTA, UNSW, USyd, UQ, MacqU, UTS                          Boeing


                                                            www.csiro.au
Aim & Capabilities



  Joint program between CIP and ICT Centre
  Aim:
     To address the science and technological issues
     associated with the future design and deployment of
     large networks (communications, sensing / monitoring,
     processing, etc.).

  Capabilities
     Sensors
     Signal processing
     Data analysis
     Communication networks
     Wireless technology
     Intelligent Algorithms


                                                     www.csiro.au
Our Approach



  Multi-Agent System

  Distributed Intelligence & Processing




  System Design Issues

  Software & Hardware Development




                                          www.csiro.au
Research Areas



  Sensors & Sensing Science

  Hardware Agents

  Multi-Agent Algorithms
     Distributed Intelligence
     Machine Learning
     Self-Assembly

  Directed Design & Control




                                www.csiro.au
Directed Design & System Issues



Vision
   “The Directed Design and Control of Multi-agent
   Systems, in the sense of being able to specify local
   agent properties to achieve desired global goals, at all
   scales from macro- to nanoscale”

      Design local agent properties to give the desired
      (emergent) global outcome
      How to herd cats and make smart dust

An important problem - common to many situations from
   macro-scale to the nano- and molecular scales.




                                                          www.csiro.au
Agents



  An AGENT has intelligence, senses, and can act
  and communicate

  An AGENT NETWORK is a lot of agents
  interacting

  AGENT NETWORKS exhibit
  EMERGENT BEHAVIOUR
    In complex systems, what “emerges” depends on
    the agent properties

  GRAND PROBLEM: Design the agents to get
  desired emergent behaviour



                                                    www.csiro.au
    Macro-scale Environments



Intelligent sensors on AAV skin       Telecommunications Networks
       locally connected




                                      Robust connectivity
Aim: Detect and analyse damage            Optimising network
     detect, assess, report, repair       efficiency
     must work even when                  Changing network
     severely damaged                     environment
                                          Dynamic re-routing
Achieve by agent design
                                          High-fault conditions

                                                          www.csiro.au
   Pseudo Meso-scale Environment



Artificial MA environment
    2-D Mesoscale cubes in liquid.
    Related to real 3-D mesoscale blocks
      • Cube faces +1, -1 or 0.
      • Opposite signs attract, like signs repel.

                                                             +/-/0
   Additional properties
    • “state” of any face can be changed
    • Small internal state machine.




                                                    +/-/0




                                                                           +/-/0
                                                             STATE
    • May sense environment                                 MACHINE
    • May change environment

Directed self-assembly                                       +/-/0

                                           Change                                  Sense

                                                                www.csiro.au
Multi-Agent Design - Basic Approach



   Emergent behaviour makes design difficult
     yet emergence gives a rich solution space
     seems to be necessary part of biological evolution

   Chosen solution
      Split the problem hierarchically
                 top-down, bottom-up (TDBU)
      • Increases solution space, makes generic design
           more likely
      • Important to keep emergent behaviour.
       •   Intermediate metrics need definition.




                                                     www.csiro.au
                Self-Assembly


                      Desired meso- or nano-structures
                      - sensors, actuators etc.



Construction


                  Constructor entities – eg. “Enzymes”


Self-assembly
                        emergent behaviour


                          Meso- or nano- “agents”

                                                         www.csiro.au
2D meso block Enzyme




                       www.csiro.au
6-pack




         www.csiro.au
Distributed Energy Management



Concept:
  deploy distributed intelligent agents to manage energy
  resources for small-to-medium enterprises (SMEs)
  SME agents will
     balance the needs and capabilities of devices, power
     consuming or power generating, throughout the SME.
     interact with other SME agents to create aggregated
     responses that are of sufficient size to be relevant in the
     National Energy Market.




                                                          www.csiro.au
      Distributed Energy Management



      Concept:
        deploy distributed intelligent agents to manage energy
        resources for small-to-medium enterprises (SMEs)
        SME agents will
           balance the needs and capabilities of devices, power
           consuming or power generating, throughout the SME.
           interact with other SME agents to create aggregated
           responses that are of sufficient size to be relevant in the
           National Energy Market.

                                                       SME
                                                               SME
NEM                  Retailer
                                                 SME
                                                         SME    SME

                                                    SME      SME

                                                                www.csiro.au
      Distributed Energy Management



      Concept:
        deploy distributed intelligent agents to manage energy
        resources for small-to-medium enterprises (SMEs)
        SME agents will
           balance the needs and capabilities of devices, power
           consuming or power generating, throughout the SME.
           interact with other SME agents to create aggregated
           responses that are of sufficient size to be relevant in the
           National Energy Market.

                                                       SME
                                                               SME
NEM                  Retailer
                                                 SME
                                                         SME    SME

                                                    SME      SME

                                                                www.csiro.au
      Distributed Energy Management



      Concept:
        deploy distributed intelligent agents to manage energy
        resources for small-to-medium enterprises (SMEs)
        SME agents will
           balance the needs and capabilities of devices, power
           consuming or power generating, throughout the SME.
           interact with other SME agents to create aggregated
           responses that are of sufficient size to be relevant in the
           National Energy Market.

                                                       SME
                                                               SME
NEM                  Retailer
                                                 SME
                                                         SME    SME

                                                    SME      SME

                                                                www.csiro.au
      Distributed Energy Management



      Concept:
        deploy distributed intelligent agents to manage energy
        resources for small-to-medium enterprises (SMEs)
        SME agents will
           balance the needs and capabilities of devices, power
           consuming or power generating, throughout the SME.
           interact with other SME agents to create aggregated
           responses that are of sufficient size to be relevant in the
           National Energy Market.

                                                       SME
                                                               SME
NEM                  Retailer
                                                 SME
                                                         SME    SME

                                                    SME      SME

                                                                www.csiro.au
      Distributed Energy Management



      Concept:
        deploy distributed intelligent agents to manage energy
        resources for small-to-medium enterprises (SMEs)
        SME agents will
           balance the needs and capabilities of devices, power
           consuming or power generating, throughout the SME.
           interact with other SME agents to create aggregated
           responses that are of sufficient size to be relevant in the
           National Energy Market.

                                                       SME
                                                               SME
                     Retailer
                                                 SME
                                                         SME    SME
NEM
                                                    SME      SME

                                                                www.csiro.au
Distributed Energy Management



  Co-investment Research with the ‘Energy’ Flagship
     End goal is reduction of greenhouse gases

  Development & Simulation of Aggregation Algorithms
     collaboration with US company – DOE funding

  Hardware at Newcastle demonstrating monitoring and
  control with varying energy loads and sources
     use ‘motes’ with wireless networking
     use X10 devices with powerline links
     interact via web interface

     coolroom environment
     office environment


                                                 www.csiro.au
HVAC Demonstrator




                    www.csiro.au
Structural Health Management (SHM)



The Vision:
   Brilliant structural systems that do not age or fail.

Concept:
  Large sensor / actuator networks
  Application to buildings, bridges, vehicles …

   Systems that monitor themselves and their environment
   Systems that can learn and adapt
   Systems that behave intelligently and can make and
   implement decisions




                                                           www.csiro.au
Ageless Aerospace Vehicle (AAV)



  Collaboration with NASA LRC
  Aim - to develop and critically examine concepts for
  integrated smart sensing and communication networks
  The ultimate goal - develop a self-monitoring, self-
  repairing aerospace vehicle.




                                                www.csiro.au
AAV



 Long-term Challenges & Issues
    Rapid non-destructive evaluation with accurate sensing
    and modelling
    A network that will function in the presence of damage

      Sensing – what, how, distributed, embedded,
      autonomous, mobile?
      Intelligence – distributed, centralised, swarm, emergent,
      multi-agent?
      Communications – network topology, bandwidth, power?
      Real-time repair – biological analogues?




                                                        www.csiro.au
AAV



What are we Doing?
  developing computational systems in which many
  autonomous software agents interact, adapt and work
  together in a common open environment, combining
  individual strategies into overall (emergent) behaviour

   We have built a hardware concept demonstrator to serve
   as an experimental test-bed

   Prime interest is in spacecraft micro-meteoroid impact
   studies




                                                    www.csiro.au
Intelligent Algorithms



  We have modelled a modular multi-cellular vehicle skin
  where each skin cell is simulated by an autonomous
  agent

  The cells also contain sensors, logic, and
  communications

  Single cells may make fast and automatic responses to
  sudden damage, while collections of cells may solve
  more complex tasks.

  For example, they may form impact boundaries




                                                  www.csiro.au
Simulation




             www.csiro.au
            Concept Demonstrator Hardware

                                      Serial Communication Links




                   400 mm
800 mm




                                                                      UDP
                                                                   Communication
                                                                      Link


           Concept Demonstrator                                               Workstation for Controlling
                                            PC Cluster for
           Hardware Containing                                                 and Monitoring Test-bed
                                           Simulating Cells
         Sensors and Physical Cells




                                                                                          www.csiro.au
     Concept Demonstrator Hardware




                             DAL




                             NAL


Acquisition Processor Link
Network Link




                                     www.csiro.au
Concept Demonstrator Hardware




                                www.csiro.au
Concept Demonstrator Hardware




                                www.csiro.au
Concept Demonstrator Hardware




                                www.csiro.au
                        Sensor Signals




                      200                                   50
                      100                                   25
                        0                                    0
Sensor signals (mV)




                      -100                                  -25
                      -200                                  -50
                          -20   -10      0        10   20      -30      -20      -10         0           10

                       20                                    10
                       10                                     5
                        0                                     0
                       -10                                   -5
                       -20
                          -30   -20      -10       0   10         -20   -10       0          10          20
                                      Time (µs)                               Time (µs)




                                                                                          www.csiro.au
Impact Boundaries




                    www.csiro.au
3D Visualisation




                   www.csiro.au
Summary



 Aim to develop local agent algorithms that will result in
 desirable emergent behaviour of the collective system
 of agents

 Simple algorithms developed so far:
    Impact boundaries.
    Damage networks.
    Clusters based on damage severity.
    Self-replication of damage region

 Entropy-based metrics developed to measure stability
 of emergent behaviour

 Strategic research coupled to application spaces

                                                   www.csiro.au

								
To top