SlideShare a Scribd company logo
Web Service Composition
          as a Planning Task


                    Experiments Using 
                 Knowledge­Based Planning 

                        Erick Martínez & Yves Lespérance

                            Department of Computer Science
                                   York University
                                   Toronto, Canada


1   Martínez & Lespérance        WSC / Knowledge-Based Planning   WPSWGS @ ICAPS-2004
Motivation
    ●   Next generation Web Services
         –   Semantic Web
         –   Reasoning / Planning
         –   Automation

    ●   Agent­oriented toolkit for advanced MAS / 
        WSC and provisioning

    ●   Previous results (knowledge­based 
        planning)
2   Martínez & Lespérance   WSC / Knowledge-Based Planning   WPSWGS @ ICAPS-2004
Automated Web Service
         Composition as Planning
    ●   WSC Problem: given a set of Web services and some 
        user defined task or goal to be achieved, automatically 
        find a composition of the available services to accom­
        plish the task.
    ●   WSC as a Planning Problem:  
         –   Predefined available services as the building blocks of a plan
         –   Many WS actions involve sensing
         –   Large search space, incomplete information in the initial 
             state
    ●   Planner that can generate conditional plans & 
        supports sensing actions
3   Martínez & Lespérance    WSC / Knowledge-Based Planning    WPSWGS @ ICAPS-2004
PKS
                 (Petrick & Bacchus, 2003)

    ●
          Generalization of STRIPS                       [Fikes & Nilsson, 1971]
           –   Four Dbs:   KF , KW , KV , KX
           –   Actions:
                      Action            Precondition                 Effects
            checkFlightSpace(n,d)    K(existsFlight(n,d))  add(Kw, availFlight(n,d))

           –   DSURs:       K(existsFlight(n,d))  ⇒  add(Kv, flightNum(n,d))
           –   Goal
           –   Planning problem: < I, A, U, G >


4       Martínez & Lespérance   WSC / Knowledge-Based Planning        WPSWGS @ ICAPS-2004
WSC in PKS

    ●   PKS primitive actions correspond to WS
         –   Knowledge­producing actions  ↔       information 
             gathering WS
         –   Physical actions  ↔   world­altering WS

    ●   New WS becomes available  →      add new 
        primitive action to domain specification


5   Martínez & Lespérance   WSC / Knowledge-Based Planning   WPSWGS @ ICAPS-2004
WS Representation in PKS

    ●   For each action Ai :
         –   encode user preferences / customization constraints 
             using desAi fluent    [McIlraith & Son, 2002]
         –   encode domain specific search control constraints us­
             ing  indAi fluent

    ●   Generic domain specification
         –   action specification (WS)

    ●   Problem specification
         –   goal + DSURs (addresses PKS limited expressiveness)

6   Martínez & Lespérance   WSC / Knowledge-Based Planning   WPSWGS @ ICAPS-2004
PKS Spec. Example
                       (Air Travel Domain)
        Action          Precondition                 Effects
     findRFlight(x)  K(airCo(x))           add(Kw, flightExists(x))
                     K(indFindRFlight(x))  add(Kf,¬indFindRFlight(x))
                     K(desFindRFlight(x))

     bookFlight(x)           K(airCo(x))                  add(Kf, bookedFlight(x))
                             K(availFlight(x))            del(Kf, availFlight(x))
                             K(indBookFlight(x))          add(Kf,¬indBookFlight(x))
                             K(desBookFlight(x))

    Domain specific update rules (DSUR)
     K(airCo(x))  ¬Kv(flightNum(x))  K(flightExists(x))  
                                            add(Kv,  flightNum(x))


                                   1 explicit parameter: company

7   Martínez & Lespérance         WSC / Knowledge-Based Planning        WPSWGS @ ICAPS-2004
PKS Goal Example
                        (Prob. BMxF)
      ●   Goal: book a flight with a price not 
          greater than the user' s maximum price

                 /* book company within budget */
                 K(x) [K(airCo(x))  K(bookedFlight(x)) 
                          K(¬priceGtMax(x))]   |
                 /* no flight booked */
                 KnowNoBudgetFlight   |
                 KnowNoAvailFlight   |
                 KnowNoFlightExists


8   Martínez & Lespérance     WSC / Knowledge-Based Planning   WPSWGS @ ICAPS-2004
User Pref. / Customization
          Constraints Example (Prob. BMxF)

      ●   DSUR 1:
           K(airCo(x))  ¬Kw(priceGtMax(x)) 
           Kv(userMaxPrice)  Kv(flightCost(x))
            add(Kw, priceGtMax(x))

      ●   DSUR 2:
           K(airCo(x))  K(¬priceGtMax(x)) 
           ¬Kw(desBookFlight(x))
            add(Kf, desBookFlight(x))

9   Martínez & Lespérance   WSC / Knowledge-Based Planning   WPSWGS @ ICAPS-2004
Experiments - Problem Set
     ●   Set of 5 problems (air travel domain): 
          –   hard constraints 
                ●   BPF: book preferred company, otherwise book any 
                    flight 
                ●   BMxF: book any flight within budget
                ●   BPMxF: book flight within budget, favour 
                    preferred company

          –   optimization constraints 
                ●   BBF: book cheapest flight
                ●   BBPF: book preferred company, otherwise book 
                    cheapest flight
10   Martínez & Lespérance    WSC / Knowledge-Based Planning   WPSWGS @ ICAPS-2004
Experimental Results with 1
       param. using DFS (in Secs.) ...
        #Co.            BPF         BMxF          BPMxF            BBF            BPBF
             2               0.00          0.00            0.00            0.02       0.10
             3               0.00          0.01            0.01            0.35       1.58
             4               0.00          0.01            0.01           53.99     259.39
             5               0.01          0.02            0.02     > tmax        > tmax
            10               0.01          0.04            0.04     > tmax        > tmax
            20               0.04          0.05            0.06     > tmax        > tmax
            50               0.31          0.51            0.60     > tmax        > tmax
           100               2.47          3.15            3.43     > tmax        > tmax

                             Results with 1 explicit parameter: company
                                     PKS v0.6­alpha­2 (Linux)
                                          (tmax = 300secs.)

11   Martínez & Lespérance       WSC / Knowledge-Based Planning               WPSWGS @ ICAPS-2004
... Experimental Results with
      5 param. using DFS (in Secs.)

         #Co.            BPF           BMxF          BPMxF             BBF              BPBF
             2                0.17           0.21           0.37            1.27             8.42
             3                0.58           0.72           1.20           24.02          109.33
             4                1.45           2.20           3.71        > tmax           > tmax
             5                3.76           4.33           4.65        > tmax           > tmax
            10               80.60          96.45         105.49        > tmax           > tmax

                                Results with 5 explicit parameters: 
                    company, origin, destination, departure date, and arrival date
                                      PKS v0.6­alpha­2 (Linux)
                                           (tmax = 300secs.)




12   Martínez & Lespérance         WSC / Knowledge-Based Planning                    WPSWGS @ ICAPS-2004
Advantages of
                              Our Approach
       ●   Modularity and re­usability

       ●   Can handle cases that previous approaches can­
           not (e.g., physical actions having direct effect on 
           sensing actions)  [McIlraith & Son, 2002]

       ●   No need for pre­specified generic plans




13   Martínez & Lespérance     WSC / Knowledge-Based Planning   WPSWGS @ ICAPS-2004
Open Problems and
                       Future Work
       ●   Customizing domain theory based on problem
            –   DSURs generation (desc. goal + user pref.)
       ●   Large search space, off­line
       ●   Optimization constraints problems do not scale up well
       ●   Representation of atomic services
       ●   Translation of OWL, DAML­S, etc. into PKS/Golog   
           specifications

       ●   IG­JADE­PKSlib toolkit
       ●   Experiments + case studies to validate performance and 
           scalability of integrated framework
       ●   Plan execution and contingency recovery

14   Martínez & Lespérance     WSC / Knowledge-Based Planning   WPSWGS @ ICAPS-2004
Thank You!



15   Martínez & Lespérance    WSC / Knowledge-Based Planning   WPSWGS @ ICAPS-2004

More Related Content

PDF
Robust Repositioning in Large-scale Networks
PPTX
Dynamic and Stochastic Berth and Quay Crane Allocation, TSL Workshop, 2013
PDF
Dynamic Dispatch Waves for Same-day Delivery
PDF
A TRAINING METHOD USING
 DNN-GUIDED LAYERWISE PRETRAINING
 FOR DEEP GAUSSIAN ...
PDF
PDF
Pr057 mask rcnn
PDF
High-Performance Physics Solver Design for Next Generation Consoles
PDF
VJAI Paper Reading#3-KDD2019-ClusterGCN
Robust Repositioning in Large-scale Networks
Dynamic and Stochastic Berth and Quay Crane Allocation, TSL Workshop, 2013
Dynamic Dispatch Waves for Same-day Delivery
A TRAINING METHOD USING
 DNN-GUIDED LAYERWISE PRETRAINING
 FOR DEEP GAUSSIAN ...
Pr057 mask rcnn
High-Performance Physics Solver Design for Next Generation Consoles
VJAI Paper Reading#3-KDD2019-ClusterGCN

What's hot (8)

PDF
Dear - 딥러닝 논문읽기 모임 김창연님
PPTX
Online Optimization Problem-1 (Online machine learning)
PPTX
Introduction of Online Machine Learning Algorithms
PDF
Demystifying Garbage Collection in Java
PDF
Learning visual representation without human label
PPTX
Tutorial on Object Detection (Faster R-CNN)
PDF
위성이미지 객체 검출 대회 - 2등
PPTX
Graphlab dunning-clustering
Dear - 딥러닝 논문읽기 모임 김창연님
Online Optimization Problem-1 (Online machine learning)
Introduction of Online Machine Learning Algorithms
Demystifying Garbage Collection in Java
Learning visual representation without human label
Tutorial on Object Detection (Faster R-CNN)
위성이미지 객체 검출 대회 - 2등
Graphlab dunning-clustering
Ad

Viewers also liked (10)

PDF
B tech ii yr i sem r9 supplementary may 2013
PPTX
Chapter 20 final 2
PPTX
Akesiu mei moana tuuholoaki
PDF
IG-JADE-PKSlib: An Agent-Based Framework for Advanced Web Service Composition...
PDF
B.tech ii year ii semester (r09) regular & supplementary april may 2013
PDF
Transit policy utsc presentation june 24 [compatibility mode]
PDF
B tech iii year ii (r07) semester supplementary
PDF
B tech iii year i (r07) semester supplementary
PPTX
Presentación1
PPT
My favorite pastimes
B tech ii yr i sem r9 supplementary may 2013
Chapter 20 final 2
Akesiu mei moana tuuholoaki
IG-JADE-PKSlib: An Agent-Based Framework for Advanced Web Service Composition...
B.tech ii year ii semester (r09) regular & supplementary april may 2013
Transit policy utsc presentation june 24 [compatibility mode]
B tech iii year ii (r07) semester supplementary
B tech iii year i (r07) semester supplementary
Presentación1
My favorite pastimes
Ad

Similar to Web Service Composition as a Planning Task: Experiments using Knowledge-Based Planning (20)

PDF
Geoff
PDF
Introducing LCS to Digital Design Verification
PDF
Data Mining
PDF
Scalabiltity in GBML, Accuracy-based Michigan Fuzzy LCS, and new Trends
PDF
The Robust Optimization of Non-Linear Requirements Models
PDF
Artigo ia traps, invariants, and dead-ends
DOC
Presentation on Machine Learning and Data Mining
PDF
Introduction to Machine Learning
PDF
Introduction to Data Mining for Newbies
PDF
The Hitchhiker’s Guide to Kaggle
PPT
Paradigm shifts in wildlife and biodiversity management through machine learning
PPTX
Chapter 2 Introduction to CR_Process.pptx
PDF
PLANNING BASED ON CLASSIFICATION BY INDUCTION GRAPH
DOC
Syllabus.doc.doc
PDF
Lecture4 - Machine Learning
PDF
Microsoft PowerPoint - Chapter01
PDF
633-600 Machine Learning
PDF
Technical Tricks of Vowpal Wabbit
PDF
Analysing and combining partial problem solutions for properly informed heuri...
DOC
MIS 542 Syllabus 08.doc
Geoff
Introducing LCS to Digital Design Verification
Data Mining
Scalabiltity in GBML, Accuracy-based Michigan Fuzzy LCS, and new Trends
The Robust Optimization of Non-Linear Requirements Models
Artigo ia traps, invariants, and dead-ends
Presentation on Machine Learning and Data Mining
Introduction to Machine Learning
Introduction to Data Mining for Newbies
The Hitchhiker’s Guide to Kaggle
Paradigm shifts in wildlife and biodiversity management through machine learning
Chapter 2 Introduction to CR_Process.pptx
PLANNING BASED ON CLASSIFICATION BY INDUCTION GRAPH
Syllabus.doc.doc
Lecture4 - Machine Learning
Microsoft PowerPoint - Chapter01
633-600 Machine Learning
Technical Tricks of Vowpal Wabbit
Analysing and combining partial problem solutions for properly informed heuri...
MIS 542 Syllabus 08.doc

Recently uploaded (20)

PDF
Unlocking AI with Model Context Protocol (MCP)
PDF
Mobile App Security Testing_ A Comprehensive Guide.pdf
PPTX
A Presentation on Artificial Intelligence
PDF
Network Security Unit 5.pdf for BCA BBA.
PDF
The Rise and Fall of 3GPP – Time for a Sabbatical?
PDF
Empathic Computing: Creating Shared Understanding
PDF
Spectral efficient network and resource selection model in 5G networks
PDF
Advanced methodologies resolving dimensionality complications for autism neur...
PDF
Architecting across the Boundaries of two Complex Domains - Healthcare & Tech...
PDF
Encapsulation_ Review paper, used for researhc scholars
PPTX
1. Introduction to Computer Programming.pptx
PDF
A comparative analysis of optical character recognition models for extracting...
PDF
cuic standard and advanced reporting.pdf
PPTX
Digital-Transformation-Roadmap-for-Companies.pptx
PDF
Encapsulation theory and applications.pdf
PPT
“AI and Expert System Decision Support & Business Intelligence Systems”
PPTX
Spectroscopy.pptx food analysis technology
PDF
Profit Center Accounting in SAP S/4HANA, S4F28 Col11
PDF
NewMind AI Weekly Chronicles - August'25-Week II
PPTX
Programs and apps: productivity, graphics, security and other tools
Unlocking AI with Model Context Protocol (MCP)
Mobile App Security Testing_ A Comprehensive Guide.pdf
A Presentation on Artificial Intelligence
Network Security Unit 5.pdf for BCA BBA.
The Rise and Fall of 3GPP – Time for a Sabbatical?
Empathic Computing: Creating Shared Understanding
Spectral efficient network and resource selection model in 5G networks
Advanced methodologies resolving dimensionality complications for autism neur...
Architecting across the Boundaries of two Complex Domains - Healthcare & Tech...
Encapsulation_ Review paper, used for researhc scholars
1. Introduction to Computer Programming.pptx
A comparative analysis of optical character recognition models for extracting...
cuic standard and advanced reporting.pdf
Digital-Transformation-Roadmap-for-Companies.pptx
Encapsulation theory and applications.pdf
“AI and Expert System Decision Support & Business Intelligence Systems”
Spectroscopy.pptx food analysis technology
Profit Center Accounting in SAP S/4HANA, S4F28 Col11
NewMind AI Weekly Chronicles - August'25-Week II
Programs and apps: productivity, graphics, security and other tools

Web Service Composition as a Planning Task: Experiments using Knowledge-Based Planning

  • 1. Web Service Composition as a Planning Task Experiments Using  Knowledge­Based Planning  Erick Martínez & Yves Lespérance Department of Computer Science York University Toronto, Canada 1 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 2. Motivation ● Next generation Web Services – Semantic Web – Reasoning / Planning – Automation ● Agent­oriented toolkit for advanced MAS /  WSC and provisioning ● Previous results (knowledge­based  planning) 2 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 3. Automated Web Service Composition as Planning ● WSC Problem: given a set of Web services and some  user defined task or goal to be achieved, automatically  find a composition of the available services to accom­ plish the task. ● WSC as a Planning Problem:   – Predefined available services as the building blocks of a plan – Many WS actions involve sensing – Large search space, incomplete information in the initial  state ● Planner that can generate conditional plans &  supports sensing actions 3 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 4. PKS (Petrick & Bacchus, 2003) ● Generalization of STRIPS  [Fikes & Nilsson, 1971] – Four Dbs:   KF , KW , KV , KX – Actions: Action Precondition Effects  checkFlightSpace(n,d)  K(existsFlight(n,d))  add(Kw, availFlight(n,d)) – DSURs: K(existsFlight(n,d))  ⇒  add(Kv, flightNum(n,d)) – Goal – Planning problem: < I, A, U, G > 4 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 5. WSC in PKS ● PKS primitive actions correspond to WS – Knowledge­producing actions  ↔    information  gathering WS – Physical actions  ↔   world­altering WS ● New WS becomes available  →    add new  primitive action to domain specification 5 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 6. WS Representation in PKS ● For each action Ai : – encode user preferences / customization constraints  using desAi fluent    [McIlraith & Son, 2002] – encode domain specific search control constraints us­ ing  indAi fluent ● Generic domain specification – action specification (WS) ● Problem specification – goal + DSURs (addresses PKS limited expressiveness) 6 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 7. PKS Spec. Example (Air Travel Domain) Action Precondition Effects  findRFlight(x)  K(airCo(x))  add(Kw, flightExists(x))  K(indFindRFlight(x))  add(Kf,¬indFindRFlight(x))  K(desFindRFlight(x))  bookFlight(x)  K(airCo(x))  add(Kf, bookedFlight(x))  K(availFlight(x))  del(Kf, availFlight(x))  K(indBookFlight(x))  add(Kf,¬indBookFlight(x))  K(desBookFlight(x)) Domain specific update rules (DSUR)  K(airCo(x))  ¬Kv(flightNum(x))  K(flightExists(x))   add(Kv,  flightNum(x)) 1 explicit parameter: company 7 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 8. PKS Goal Example (Prob. BMxF) ● Goal: book a flight with a price not  greater than the user' s maximum price /* book company within budget */ K(x) [K(airCo(x))  K(bookedFlight(x))           K(¬priceGtMax(x))]   | /* no flight booked */ KnowNoBudgetFlight   | KnowNoAvailFlight   | KnowNoFlightExists 8 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 9. User Pref. / Customization Constraints Example (Prob. BMxF) ● DSUR 1: K(airCo(x))  ¬Kw(priceGtMax(x))  Kv(userMaxPrice)  Kv(flightCost(x))  add(Kw, priceGtMax(x)) ● DSUR 2: K(airCo(x))  K(¬priceGtMax(x))  ¬Kw(desBookFlight(x))  add(Kf, desBookFlight(x)) 9 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 10. Experiments - Problem Set ● Set of 5 problems (air travel domain):  – hard constraints  ● BPF: book preferred company, otherwise book any  flight  ● BMxF: book any flight within budget ● BPMxF: book flight within budget, favour  preferred company – optimization constraints  ● BBF: book cheapest flight ● BBPF: book preferred company, otherwise book  cheapest flight 10 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 11. Experimental Results with 1 param. using DFS (in Secs.) ... #Co. BPF BMxF BPMxF BBF BPBF 2 0.00 0.00 0.00 0.02 0.10 3 0.00 0.01 0.01 0.35 1.58 4 0.00 0.01 0.01 53.99 259.39 5 0.01 0.02 0.02 > tmax > tmax 10 0.01 0.04 0.04 > tmax > tmax 20 0.04 0.05 0.06 > tmax > tmax 50 0.31 0.51 0.60 > tmax > tmax 100 2.47 3.15 3.43 > tmax > tmax Results with 1 explicit parameter: company PKS v0.6­alpha­2 (Linux)  (tmax = 300secs.) 11 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 12. ... Experimental Results with 5 param. using DFS (in Secs.) #Co. BPF BMxF BPMxF BBF BPBF 2 0.17 0.21 0.37 1.27 8.42 3 0.58 0.72 1.20 24.02 109.33 4 1.45 2.20 3.71 > tmax > tmax 5 3.76 4.33 4.65 > tmax > tmax 10 80.60 96.45 105.49 > tmax > tmax Results with 5 explicit parameters:  company, origin, destination, departure date, and arrival date PKS v0.6­alpha­2 (Linux)  (tmax = 300secs.) 12 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 13. Advantages of Our Approach ● Modularity and re­usability ● Can handle cases that previous approaches can­ not (e.g., physical actions having direct effect on  sensing actions)  [McIlraith & Son, 2002] ● No need for pre­specified generic plans 13 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 14. Open Problems and Future Work ● Customizing domain theory based on problem – DSURs generation (desc. goal + user pref.) ● Large search space, off­line ● Optimization constraints problems do not scale up well ● Representation of atomic services ● Translation of OWL, DAML­S, etc. into PKS/Golog    specifications ● IG­JADE­PKSlib toolkit ● Experiments + case studies to validate performance and  scalability of integrated framework ● Plan execution and contingency recovery 14 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004
  • 15. Thank You! 15 Martínez & Lespérance WSC / Knowledge-Based Planning WPSWGS @ ICAPS-2004