Skip to content. | Skip to navigation

Personal tools
Sections

Florian Marquardt (2008)

Synthesis of Assistant Services in Adhoc Multi Media Environments

In: Proceedings des gemeinsamen Workshops der Graduiertenkollegs 2008, ed. by Malte Diehl and Henrik Lipskoch and Roland Meyer and Christian Storm, pp. 164, Schloss Dagstuhl, GITO-Verlag Berlin.

In the context of project MuSAMA ad hoc ensembles are understood as spontaneously forming communities of various multi-media devices in a spatially confined area. They are characterized by dynamic devices that offer individual services to the user in an unintrusive manner. The goal of this research is to provide smart user assistance in such environments by service composition. This user assistance can be distinguished in four steps: Identifying of user intentions, deriving goals, synthesizing strategies and finally the execution of the strategy. Composition of services takes places in the third phase (strategy synthesis). Each composition method used during the synthesis phase aims at combining existing services to fulfill the identified goal. Many different methods for composing services exists, they are as different as the smart environments in which they shall work. Therefore, methods are required to select promising strategies for the smart environment at hand. Using background knowledge and sensor data, intentions of persons located in the room can be derived. Based on the intentions of the persons and the capabilities of the devices, suitable services shall be identified and offered. For more advanced services often the cooperation of devices is required which is based on the composition of services. The most common used approaches for service composition are AI planning or workflow based. Selecting a suitable method and strategy for composition depends on the type and structure of the environment. Benchmarks and even metrics are missed that help to characterize smart environments. Having this benchmarks allow to distinguish among the different service composition strategies. The challenge of a quantitative analysis of different composition methods in smart environments is still unsolved. Also in the semantic web service community a lack of methods for evaluation is observed. To this end we identified several metrics. Beginning with simple characteristics like the number of devices or number of services, we also identified more complex characteristics like structural homogeneity of devices or the interdependencies of the services. Apart from this quantitative characterizations we also identified qualitative characterizations of the structure and behavior of smart environments, as for example the decision when a currently working planning process must be interrupted, or how and especially when to handle state changes of the environment. Knowing the mentioned metrics and knowing furthermore their impact on the composition process in smart environments, enables to automatically choose the best possible composition approach for a given environment. Whereas ''best possible'' depends on the context of the environment, such as time and space. The behavior of smart environment which should be measured using our metrics is investigated by computer simulation. Therefore we designed and implemented a general and abstract model of smart environment. We included key characteristics that likely have an effect on the behavior of the different strategies. For implementation we used the PDEVS formalism and the simulation system JAMES II. The architecture we modeled is central. It consists of devices, which carry one or more services, a repository, an intention analysis and a composer. All participating services register themselves at the repository. For this reason the address of the repository is known by all services. Furthermore an intention analysis is included in our model. It sends desired goals to the composer, expecting that the composer returns a plan that fulfills the goals. The composer is informed by the repository every time a change in the structure of the environmentAs structure we define the set of services in the environment. takes place as well as a new goal is derived by the intention analysis. Both cases results in a new planning process at the composer. In our model we implemented some basic behavior. As we use DEVS, that is a state based formalism, our models have states. First evaluations that focus on planning strategies for composition show that making runtime expectations is not straight forward and thus give a hint that in the smart environment we are currently developing different composition strategies has to be further explored. If new devices enter into the ensemble the services they offer have to be ``understood'', and their interfaces as well as their semantics must be understandable to other services in the ensemble. Our future efforts will be based on standards, e.g. WSDL for interface definition and OWL-S for semantic description of the services.

Document Actions