Pervasive computing is an effort towards achieving the vision of an invisible computing fabric around us [1, 2]. For effective management of such pervasive systems we need a way to capture the dynamic environment around us and provide adaptive systems that proactively tailor themselves depending upon specific contexts such as location, time, and other parameters. In other words, individual services are embedded ubiquitously in such pervasive systems and the need is to compose them so as to achieve specific user requirements. We find that Service-Oriented Architecture (SOA) is particularly appropriate for modelling such pervasive workflows. Web service composition in the SOA paradigm is a widely researched issue for the past few years which have introduced new thoughts and approaches to this problem. Dustdar et al. [8] reveals the intrinsic problems, complexities and limitations of the topic. From the perspective of the user we can classify the problem of service composition into two broad categories: (i) task-based and (ii) event-based. Most previous works have focused on the former assuming a given "task-template" [3, 4]. Service composition engines reformat user-requests into abstract well-defined task templates that are framed as a graph of abstract services. These abstract services are the sub goals that constitute a global goal to be achieved in the best way through a service composition process. However, a task-template requires prior understanding of the specific input and output requirements of each of the abstract sub services. In a dynamic pervasive environment such prior understanding is not always possible owing to the characteristic uncertainty of the tasks themselves and hence they cannot be modeled as a goal for a service composition process.
展开▼
