Agent-Based Information Sharing for Ambient Intelligence - - PowerPoint PPT Presentation

agent based information sharing for ambient
SMART_READER_LITE
LIVE PREVIEW

Agent-Based Information Sharing for Ambient Intelligence - - PowerPoint PPT Presentation

Agent-Based Information Sharing for Ambient Intelligence Andrei Olaru AI-MAS Group, University Politehnica Bucharest LIP6,


slide-1
SLIDE 1
  • Agent-Based

Information Sharing for Ambient Intelligence

——————————————————————— Andrei Olaru AI-MAS Group, University Politehnica Bucharest LIP6, University Pierre et Marie Curie, Paris 09.07.2010

1/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-2
SLIDE 2

Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

Agent-Based Information Sharing for Ambient Intelligence

——————————————-

  • verview

2/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-3
SLIDE 3
  • Agent-Based Information

Sharing for Ambient Intelligence What is AmI? Layers Sharing Agents Application Context Scenario Results Conclusions

Ambient Intelligence –

  • r

AmI – is an ubiquitous electronic environment that supports people in their daily tasks, in a proactive, but ”invisible” and non-intrusive manner.[Ramos et al., 2008, Weiser, 1993]

3/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-4
SLIDE 4
  • Agent-Based Information

Sharing for Ambient Intelligence What is AmI? Layers Sharing Agents Application Context Scenario Results Conclusions

Ambient Intelligence –

  • r

AmI – is an ubiquitous electronic environment that supports people in their daily tasks, in a proactive, but ”invisible” and non-intrusive manner.[Ramos et al., 2008, Weiser, 1993] People

3/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-5
SLIDE 5
  • Agent-Based Information

Sharing for Ambient Intelligence What is AmI? Layers Sharing Agents Application Context Scenario Results Conclusions

Ambient Intelligence –

  • r

AmI – is an ubiquitous electronic environment that supports people in their daily tasks, in a proactive, but ”invisible” and non-intrusive manner.[Ramos et al., 2008, Weiser, 1993] People · Devices

3/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-6
SLIDE 6
  • Agent-Based Information

Sharing for Ambient Intelligence What is AmI? Layers Sharing Agents Application Context Scenario Results Conclusions

Ambient Intelligence –

  • r

AmI – is an ubiquitous electronic environment that supports people in their daily tasks, in a proactive, but ”invisible” and non-intrusive manner.[Ramos et al., 2008, Weiser, 1993] People · Devices · Communication

3/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-7
SLIDE 7
  • Agent-Based Information

Sharing for Ambient Intelligence What is AmI? Layers Sharing Agents Application Context Scenario Results Conclusions

Ambient Intelligence –

  • r

AmI – is an ubiquitous electronic environment that supports people in their daily tasks, in a proactive, but ”invisible” and non-intrusive manner.[Ramos et al., 2008, Weiser, 1993] People · Devices · Communication Problem: How to get the relevant information to the interested users?

3/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-8
SLIDE 8
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction A Layered Perspective Sharing Agents Application Context Scenario Results Conclusions

(layers based on [Seghrouchni, 2008])

4/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-9
SLIDE 9
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction A Layered Perspective Sharing Agents Application Context Scenario Results Conclusions

(layers based on [Seghrouchni, 2008])

Hardware

4/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-10
SLIDE 10
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction A Layered Perspective Sharing Agents Application Context Scenario Results Conclusions

(layers based on [Seghrouchni, 2008])

Hardware · Network

4/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-11
SLIDE 11
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction A Layered Perspective Sharing Agents Application Context Scenario Results Conclusions

(layers based on [Seghrouchni, 2008])

Hardware · Network · Interoperability

4/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-12
SLIDE 12
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction A Layered Perspective Sharing Agents Application Context Scenario Results Conclusions

(layers based on [Seghrouchni, 2008])

Hardware · Network · Interoperability · Application

4/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-13
SLIDE 13
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction A Layered Perspective Sharing Agents Application Context Scenario Results Conclusions

(layers based on [Seghrouchni, 2008])

Hardware · Network · Interoperability · Application · Interface

4/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-14
SLIDE 14
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction A Layered Perspective Sharing Agents Application Context Scenario Results Conclusions

(layers based on [Seghrouchni, 2008])

Hardware · Network · Interoperability · Application · Interface

4/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-15
SLIDE 15
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Information Sharing Agents Application Context Scenario Results Conclusions

· The users must get the information that is interesting to them. → context-awareness is needed, to calculate relevance.

5/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-16
SLIDE 16
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Information Sharing Agents Application Context Scenario Results Conclusions

· The users must get the information that is interesting to them. → context-awareness is needed, to calculate relevance. · Ambient intelligence must be reliable and dependable. → distribution is absolutely necessary.

5/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-17
SLIDE 17
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Information Sharing Agents Application Context Scenario Results Conclusions

· The users must get the information that is interesting to them. → context-awareness is needed, to calculate relevance. · Ambient intelligence must be reliable and dependable. → distribution is absolutely necessary. Our goal: build a multi-agent system for the context-aware sharing of information.

5/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-18
SLIDE 18
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Why Agents? Application Context Scenario Results Conclusions

· Agents satisfy the needs of AmI in terms of: · autonomy · reactivity · proactivity · planning · reasoning · anticipation

6/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-19
SLIDE 19
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Why Agents? Application Context Scenario Results Conclusions

· Agents satisfy the needs of AmI in terms of: · autonomy · reactivity · proactivity · planning · reasoning · anticipation · Agents also offer beliefs, goals, intentions and easier implementation of a human-inspired behaviour.

6/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-20
SLIDE 20
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Why Agents? Application Context Scenario Results Conclusions

· Agents satisfy the needs of AmI in terms of: · autonomy · reactivity · proactivity · planning · reasoning · anticipation · Agents also offer beliefs, goals, intentions and easier implementation of a human-inspired behaviour. · Agents can provide the intelligent component of Ambient Intelligence – they are distributed, they act locally, etc.

[Ramos et al., 2008] 6/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-21
SLIDE 21
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

Our goal: build a multi-agent system for the context-aware sharing of information. · how can we obtain context-aware behaviour with simple agents acting locally? · features:

◮ local behaviour ◮ simple behaviour ◮ small knowledge base ◮ use feedback and self-organization techniques ◮ use simple and generic measures for context-awareness

7/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-22
SLIDE 22
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context-Awareness Scenario Results Conclusions

The measures of context-awareness are directed at local information sharing based on importance, relatedness to domains of interest, and validity in time.

◮ space-locality – the information spreads around its

source

◮ pressure – translates directly into relevance of the

information – controls how fast the information spreads.

◮ specialty – specifies to which domains of interest the

information is related – controls the direction of the spread.

◮ persistence – specifies for how long the information is

valid – controls the time for which the information will remain in the system.

8/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-23
SLIDE 23
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context-Awareness Scenario Results Conclusions

The measures of context-awareness are directed at local information sharing based on importance, relatedness to domains of interest, and validity in time.

◮ space-locality – the information spreads around its

source

◮ pressure – translates directly into relevance of the

information – controls how fast the information spreads.

◮ specialty – specifies to which domains of interest the

information is related – controls the direction of the spread.

◮ persistence – specifies for how long the information is

valid – controls the time for which the information will remain in the system. These measures are aggregated into a measure of relevance.

8/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-24
SLIDE 24
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Application Scenario Results Conclusions 9/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-25
SLIDE 25
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Application Scenario Results Conclusions

· create a certain distribution of interest – by inserting facts with low persistence and pressure, and different specialties.

9/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-26
SLIDE 26
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Application Scenario Results Conclusions

· create a certain distribution of interest – by inserting facts with low persistence and pressure, and different specialties. · test the behaviour of the system by inserting 3 data facts,

  • f

different specialty, with medium pressure and high persistence.

9/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-27
SLIDE 27
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Application Scenario Results Conclusions

· create a certain distribution of interest – by inserting facts with low persistence and pressure, and different specialties. · test the behaviour of the system by inserting 3 data facts,

  • f

different specialty, with medium pressure and high persistence. · test the behaviour of the system by inserting 1 data fact with high pressure.

9/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-28
SLIDE 28
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Application Scenario Results Conclusions

· create a certain distribution of interest – by inserting facts with low persistence and pressure, and different specialties. · test the behaviour of the system by inserting 3 data facts,

  • f

different specialty, with medium pressure and high persistence. · test the behaviour of the system by inserting 1 data fact with high pressure. Expect: control of the resulting distributions depending

  • n their respective measures of context-awareness.

9/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-29
SLIDE 29
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

fact distribution in the agent grid agents: specialty specialty for each domain pressure

10/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-30
SLIDE 30
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

fact distribution in the agent grid agents: specialty specialty for each domain pressure

10/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-31
SLIDE 31
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

fact distribution in the agent grid agents: specialty specialty for each domain pressure

10/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-32
SLIDE 32
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

fact distribution in the agent grid agents: specialty specialty for each domain pressure

10/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-33
SLIDE 33
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

fact distribution in the agent grid agents: specialty specialty for each domain pressure

10/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-34
SLIDE 34
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

fact distribution in the agent grid agents: specialty specialty for each domain pressure

10/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-35
SLIDE 35
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

fact distribution in the agent grid agents: specialty specialty for each domain pressure

10/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-36
SLIDE 36
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

fact distribution in the agent grid agents: specialty specialty for each domain pressure distribution of high-pressure fact

10/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-37
SLIDE 37
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

fact distribution in the agent grid agents: specialty specialty for each domain pressure distribution of high-pressure fact

10/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-38
SLIDE 38
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

fact distribution in the agent grid agents: specialty specialty for each domain pressure distribution of high-pressure fact

10/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-39
SLIDE 39
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

Why obtaining these results is not straightforward:

◮ agents only know about 20 facts, only few of them

being about their neighbours.

◮ agents are both pro-active and reactive, so feedback

may generate overloads in their message inbox.

◮ knowledge bases are very limited in size, so it is

essential to have a good algorithm to sort knowledge and forget irrelevant knowledge.

11/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-40
SLIDE 40
  • Agent-Based Information

Sharing for Ambient Intelligence Introduction Layers Sharing Agents Application Context Scenario Results Conclusions

◮ a multi-agent system was built, with agents that have

local knowledge and interact locally.

◮ simple measures for context-awareness were developed,

that allow the calculus of relevance of facts, in function

  • f their context, and the agent’s context.

◮ the system was tested and relevant results were

  • btained.

12/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-41
SLIDE 41
  • Ramos, C., Augusto, J., and Shapiro, D. (2008).

Ambient intelligence - the next step for artificial intelligence. IEEE Intelligent Systems, 23(2):15–18. Seghrouchni, A. E. F. (2008). Intelligence ambiante, les defis scientifiques. presentation, Colloque Intelligence Ambiante, Forum Atena. Weiser, M. (1993). Some computer science issues in ubiquitous computing. Communications - ACM, pages 74–87. 13/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-42
SLIDE 42
  • 13/ 14

Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010

slide-43
SLIDE 43
  • Thank you!

——————————————————————— Any Questions?

14/ 14 Computer Science & Engineering Department . . Andrei Olaru . SEE-MAS Summer School . Bucharest, 09.07.2010