Preference-based AI Planning for Web Service Composition

Abstract

A web service is a technology that allows data to beexchanged between applications using open protocols andstandards. Web services generally offer basic functions that solvespecific problems. The composition of services occurs when morecomplex problems need to be solved using successive invocationsto other services. The automatic composition of web services is acurrent problem. Due to the continuous creation and updating ofservices, the manual composition of new services becomesimpractical. Planning algorithms are an alternative to automatethe process of web services composition. In particular, theplanning algorithms with preferences allow users to guide thealgorithm to find a solution in line with their requirements (e.g.prioritizing the use of services according to properties such asavailability, cost, response time, etc.). This paper introduces aPlanning approach with preferences to compose and run complexservices from a set of web services given as input. To evaluate theapproach, different preferences about the input services wereused to obtain composition of services. As a result, we measuredthe feasibility of the proposed approach and the impact of the useof preferences on the composition result.