Adaptive Composite Service Plans for Ubiquitous Computing

Ubiquitous Computing environments require dynamic adaptable application and network services because of their need to adapt to rapidly change context e.g. new users and devices being introduced, changes in connectivity, changes in user needs and requirements, etc. To support such environments, management systems need to dynamically adapt service execution and behaviour. One of the approaches being suggested currently is that of AI planning, where execution plans (sequences of service executions) can be generated dynamically. However, they often have being criticised for being inflexible since it only adapts to changes in context through re-planning. This paper explores an approach to make service plans more adaptable to changing context prior to as well as during execution time without re-planning. In our approach the service’s non-functional adaptive behaviour is explicitly documented by a service developer in a run-time accessible model. This paper proposes the use of Finite State Machines (FSM) to describe the adaptive behaviour of a particular service and the formation of a composite FSM representing the adaptive behaviour of the constituent services. The paper then proposes policies to be used as rules by which adaptive behaviours can be specified prior to, or even during, execution. For a composite service, these policies are refined and applied to its constituent services, thus facilitating automatic refinement of high level policies to lower level policies executable appropriate constituent service. The paper presents a case study to illustrate the outlined approach and identify the strengths and challenges in applying this approach.