论文阅读 A Comparative Illustration of AI Planningbased Web Services Composition

Title

A Comparative Illustration of AI Planning-based Web Services Composition

Journal

ACM SIGecom Exchanges

Year

2005

Author

Seog-Chan Oh, Dongwon Lee

Level

Introductory

Comment

根据自动/手动，复杂性，规模对WSC问题进行区分，然后讲述了AI方法对各种不同的WSC问题的适用性，重点研究了AI Planning方法对一般WSC问题的应用。

 

1. 文章结构

Propose the Problem Setting: Find the Restaurant and give the concept of WSC

Classification of WSC Problem

Manual vs. Automatic Workflow Composition

Simple vs. Complex Operator

Small vs. LargeScale

Overview of Matching Approaches

Approach-1: exact match using syntactic equivalence

Approach-2: approximate match using distance functions

Approach-3: semantic match using ontologies

AI-Planning based algorithms for WSC problem

Graphplan based planning

SATPlan based reduction

Integer Linear Programming formulation

Conclusion

2. Claims

Web services based e-service workflow problem can be formulated as AI-planning problem

3. WSC problem classification

(1) Manual vs. Automatic Workflow Composition

Manual approach is not appropriate for large-scale WSC problem.

(2) Simple vs. Complex Operator

Simple WSC: sequential AND Composition

(3) Small vs. Large Scale

Small-scale WSC problem: exhaustive search algorithms

Large-scale problem: approximate algorithms

The automatic composition approach can be complementary to the manual approach such that a few feasible workflows generated from the automatic approach are in turn presented to domain experts who may choose one of them, and refine it further manually.

4. Different WSC problem and the corresponding AI method

(1) WSC without complex operator:

heuristic sub-optimal algorithms (e.g. A *)

(2) Operators are complex and some specific constraint rules must be checked

rule-based expert systems

(3) More General WSC problem

AI planning methods

5. AI-Planning based algorithms for WSC problem

Graphplan and ILP are suitable for planning problem with complex operators in a small-scale.

SATPlan can be used to find sub-optimal compositions for a large-scale problem with complex operator

Appendix

A.1 Definition of WSC (Web Service Composition) Problem

In order to fully satisfy a request r, one has to compose multiple web services, {w1, w2, …, wn } in sequential or parallel way such that:

(1) ∈{w1,w2,…,wn}, wiⁱⁿ can be grounded when wi^out is required at a particular state in composition

(2) ⁱⁿ Uw1^out … Uwn^out) ⊇r^out

 

A.2 Background

(1) The general WSC problem can be reduced to the satisfiabiliy problem[Vossen et al. 1999].