jaaf: a framework to implement self-adaptive agents able ...€¦ · e jaaf: a framework to...
TRANSCRIPT
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
JAAF: A Framework to Implement Self-Adaptive Agents Able to Deal with Web Services
Baldoino [email protected]
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
11/03/09 2
Roadmap
• Motivation
• OWL-S
• Model– Class Diagram
• Case Study : – GeoRisc
– Adaptive Personal Web Page
• Related Work
• Conclusion
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Motivation
• Service-oriented computing (SOC) has taken hold in business in, for instance, the use of shipping services in e-commerce transactions; the aggregation of hotel, car rental, and airline services.
• Therefore, it is necessary to provide techniques to discover, invoke, compose and monitor web services.
• Semantic Web Service (SWS) has been pointed as a way to address these issues.
• Although SWS can solve some of the mentioned issues, the complexity of current systems has directed the software engineering community to look for systems able to adjust or adapt their behavior in response to requirement changes.
3
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Motivation
• Considering that adaptive agents present properties like: reasoning, learning, autonomy and pro-activity, multi-agent system is a paradigm that fits on these concerns.
• The Java self-Adaptive Agent Framework (JAAF) was proposed.
– It extends the JADE framework;
– It offers support to the implementation of different self-adaptation processes composed of activities that can perform:
• Collect of data;
• Analysis;
• Decisions.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Motivation
• The Java self-Adaptive Agent Framework (JAAF)
– It provides reasoning mechanisms based on:
• Rules;
• Cases;
• Genetic Algorithm.
– It offers selection mechanisms based on:
• Utility Function;
• Reputation.
– It provides flexibility to create different plans for self-adaptation.
5
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
OWL-S
6
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Framework Model – Class Diagram
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
JAAF: Collect
• It is responsible for providing mechanisms to collect, aggregate and filter (format) data collected from the application.
• The collect has two sub-activities:
– Sensor: It defines the place where the data should be collected (database, log, etc).
– Format: It defines the format of the collected data.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
JAAF: Analyze
• The analyze activity is responsible for providing mechanisms that analyze the data collected in the previous activity in order to detect problems and suggest new solutions.
• This activity gives support to three techniques:
– Rule-based reasoning (forward chaining, backward chaining and fuzzy logic) [12] ;
– Case-based reasoning [8] ;
– Genetic algorithm [11].
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
JAAF: Decision and Effector
• Decision
– Decision is the activity responsible for deciding which action (or behavior) will be the next one to be executed by the agent, while trying to achieve the goal.
– This activity gives support to two techniques:
• Reputation;
• Utility Function.
• Effector
– It receives the selected action from the Decision activity, and informs the agent the action to be executed.
– When the action is executed, the control loop can be executed again whether any self-adaptation is necessary.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
JAAF: Hot-spots and Frozen-spots
• The hot-spots specifically defined in JAAF are:
– Agent (AdaptationAgent class):
• By extending such class and implementing the executedPlan method, it is possible to define different algorithms to execute the plans of an agent.
– Plan of self-adaptation (PlanAdaptation class):
• It is possible to define new control loops (or plans) and the sequence to execute the activities of the control loops. JAAF already provides a default control loop implemented in the ControlLoop class.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
JAAF: Hot-spots and Frozen-spots
– Activities (Behaviour class): • It is possible to define new activities to be called by the control
loops by extending the Behaviour class. JAAF already offers four activities (Collect, Analyze, Decision and Effector).
– Sensor (Sensor class):• One can define when and where the data should be collected.
– Format (Format class): • It is possible to define the format of the data to be collected by the
Sensor.
– Selection techniques:• Reputation and Utility Function.
– Intelligent Algorithm module: • JAAF offers three kinds of algorithms: rule-based reasoning
(forward chaining, backward chaining and fuzzy logic), case-based reasoning and genetic algorithm.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
GeoRisc
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
© LES/PUC-Rio
Problema
BR-101 – Rio-Santos
Quitandinha, Petrópolis, Dezembro de 2001
Morin, Petrópolis, Dezembro de 2002
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
© LES/PUC-Rio
Camadas
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
© LES/PUC-Rio
Agente de Recomendação
Tipos de análises Modelo Características Dados necessários
Escalas recomendadasEscalas recomendadasEscalas recomendadasTipos de análises Modelo Características Dados necessários
Regional Média Grande
Inventário
Distribuição dos escorregament
os
Análise da distribuição e classificação
3 sim sim sim
InventárioAtividade dos
escorregamentos
Análise das mudanças do padrão temporal
4,5,14,15,16,17 não sim simInventário
Densidade dos escorregament
os
Cálculo da densidade nas unidades de terreno
ou no mapa de isopletas
1,2,3 sim não não
Heurísticos
GeomorfológicaUsa observações de
campo do perito no zoneamento
2,3,4 sim sim sim
Heurísticos
Combinação qualitativa
Perito fornece ponderação de
valores dos mapas
2,3,5,6,7,8,9,10
,12,14,16,18sim sim não
Estatísticas
Bivariada Calcula a importância da combinação de
fatores
2,3,5,6,7,8,9,10
12,14,16,18não sim não
Estatísticas
MultivariadaCalcula a formula de
previsão a partir de uma matriz de
dados
2,3,5,6,7,8,9,10
12,14,16,18não sim não
Determinísticas
Fator de segurançaAplicada a modelos de
hidrologia e estabilidade de
taludes
6,11,12,13,16
20,21,22,23não não sim
• Regras
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
© LES/PUC-Rio
Regras Iniciais
Geomorfologia
1. Mapeamento de unidades de terreno
2. Unidades e subunidades geomorfológicas
3. Escorregamentos recentes
4. Escorregamentos antigosTopografia
5. Modelo digital do terreno
6. Mapa de declividades
7. Mapa de direção dos taludes
8. Comprimento do talude
9. Concavidade / convexidade
Geotecnia
10. Litologia
11. Pedologia
12. Mapa geológico estrutural
13. Acelerações sísmicas
Uso do Solo
14. Infra-estrutura recente
15. Infra-estrutura antiga
16. Mapa de uso do solo recente
17. Mapa do uso do solo antigo
Hidrologia
18. Drenagem
19. Bacia hidrológica
20. Regime de chuvas
21. Temperatura
22. Evapotranspiração
23. Mapa do Nível de água
Caracterização Escalas Áreas mínimas de estudo
Nacional 1:1.000.000 250.000 km2
Regional 1:100.000 a 1:500.00
0
2.500 a 62.500
km2Média 1:25.000 a
1:50.00056,25 a 625
km2
Grande 1:5.000 a 1:15.000
6,25 a 56,25 km2
Local 1:5.000 e maiores
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Case Study: Motivation
• Landslides are natural phenomena, which are difficult to predict since they depend on many (unpredicted) factors and on relationships among those factors.
• One of the main challenges faced by specialists is to decide the most appropriate configuration of susceptibility model to generate susceptibility maps (SM).
• In this context, we used the JAAF framework to create a multi-agent system in order to generate a SM that shows the places with landslide risks of Rio de Janeiro, a city in Brazil.
• Each application agent is able to adapt the configuration of its susceptibility model in order to meet the SM that represents the reality.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Vision
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Manager Agent
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Service Agent
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Adaptive Personal Web Pages
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Vision
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Related Works: Rainbow
David Garlan, Shang-Wen Cheng, An-Cheng Huang, Bradley Schmerl and Peter Steenkiste. Rainbow: Architecture-Based Self-Adaptation with Reusable Infrastructure. In IEEE Computer, Vol. 37(10), October 2004.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Related Works: A framework for Dynamic Adaptation of Power-Aware Server Clusters
PETRUCCI, V. ; LOQUES, O. ; Mossé, D. A framework for dynamic adaptation of power-aware server clusters. In: 24th ACM Symposium on Applied Computing, 2009, Honolulu, Hawaii, USA. SAC '09: Proceedings of the 24th ACM Symposium on Applied Computing. New York, NY, USA : ACM, 2009. p. 1034-1039.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Related Works: DySOA
Bosloper, I.; Silgee, J.; Nijhuis, J. and Hammer, D., 2005, Creating Self-Adaptive Service Systems with DySOA , Proceedings of the Third European Conference on Web Services.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Related Works: DySOA
Poggi, A.; Tomaiuolo, M. and Turci, P., 2007, An Agent-Based Service Oriented Architecture, Agenti e Industria: Applicazioni tecnologiche degli agenti software.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Conclusion and Future Works
• New control-loops of self-adaptation
• Safe Adaptation
– Norms
• Software Test and Metrics
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Reference
1. S. Dobson, S. Denazis, A. Fernández, D. Gaiti, E. Gelenbe, F. Massacci, P. Nixon, F. Saffre, N. Schmidt, and F. Zambonelli. A survey of autonomic communications. ACM Transactions Autonomous Adaptive Systems (TAAS), 1(2):223-259, December 2006.
2. David Garlan, Shang-Wen Cheng, An-Cheng Huang, Bradley Schmerl and Peter Steenkiste. Rainbow: Architecture-Based Self Adaptation with Reusable Infrastructure. In IEEE Computer, Vol. 37(10), October 2004.
3. Kaiser, G.; Parekh, J.; Gross, P. and Valetto, G. Kinesthetics eXtreme: An external infrastructure for monitoring distributed legacy systems. In Proceedings of the Autonomic Computing Workshop at the5th Annual International Workshop on Active Middleware Services (AMS), 2003.
4. Bigus, J. P.; Schlosnagle, D. A., Pilgrim, J. R.; et. al..ABLE: A toolkit for building multiagent autonomic systems. IBM Syst. J. 41, 3, 350–371, 2002.
5. Jennings, N. R. and Wooldridge, “M. Agent-oriented software engineering,” In Bradshaw, J. (Ed.) Handbook of Agent Technology, AAAI/MIT Press, 2000.
6. Wooldridge, M. and Jennings, “N. R. Pitfalls of agent-oriented development,” Proceedings of the Second International Conference on Autonomous Agents (Agents'98), ACM Press, pp. 385-391, 1998.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
Reference
1. Huebscher, M. C. and McCann, J. A. A survey of Autonomic Computing—Degrees, Models, and Applications. ACM Computing Survey, August 2008.
2. A. Amodt and E. Plaza. Case-based reasoning: Foundational issues, methodological variations, and system approaches. In AI Communications, volume 7:1, pages 39–59. IOS Press, March 1994.
3. Melanie Mitchell, An Introduction to Genetic Algorithms (Complex Adaptive Systems), The MIT Press, February 6, 1998.
4. Costa, A., Lucena, C. J. P.; Silva, V., Cowan, D.; Alencar, P., A Hybrid Diagnostic-Recommendation System for Agent Execution in Multi-Agent Systems, ICSOFT 2008 – 3rd International Conference on Software and Data Technologies, Porto, Portugal, July 2008.
Wednesday, November 4, 2009
La
bo
rató
rio
de
En
ge
nh
ari
a d
e S
oft
wa
re
The End!!
Questions?
Wednesday, November 4, 2009