Information Technology Reference
In-Depth Information
The functionality of SHOP2 consists of three basic steps. In the first one, the do-
main is constructed by the process OWL-S files of the available web services. The
atomic services are represented by operators and methods for analyzing the complex
services to simpler ones are constructed. In the second step, the composition problem
is transformed to planning problem. This is realized by describing the problem as an
abstract composite process that need decomposition with the use of methods so as to
obtain simple processes that refer to web services. In the third step, the problem is
solved by decomposing the tasks and creating the plan, i.e. is the description of the
final composite service.
Another technique, analyzed in [12], is based on situation calculus, where the
states are not considered as instances of the environment but as sequences of actions
that were executed in the past and resulted to this state. This technique uses also the
language Golog (alGOL in LOGic), which is based on logic and the problems that are
encoded in it can be solved by methods that use logic. For the appropriate representa-
tion of the planning problem in Golog, the language was extended so as to be able to
contain constraints on the composition process defined by the user. These constraints
in essence reflect the desired outputs. The OWL-S descriptions are used as require-
ments of the processes that must be executed and also as descriptions of the actions
that are provided by the web services. The composition problem is transformed into
the problem of finding the appropriate Golog program that when executed, all the
defined constraints will be satisfied. In the solution process, intelligent agents are
used whom knowledge base contains the preconditions and the results of the services,
encoded in situation calculus terms. The available web services correspond to opera-
tors, primitive or composite. The role of the agents is the inference on the web servic-
es, in order to discover, execute and compose them.
A different and quite simple web services composition method is presented in [18].
It is based on regression in a state space. The algorithms belonging to this category
start from exploring the goals that must be succeeded and seek for the actions that
lead from the goals to the initial state. The method proposed introduces a new struc-
ture called SLM (Semantic Links Matrix) and is a table containing the values of
semantic relevance between the parameters of the web services. For the construction
of this table, the process models and the relative ontologies of the atomic services are
used. Generally, the SLM structure groups the candidate web services based on their
semantic relevance and in the same time provides information on their quality charac-
teristics so as to ease the choice among them. The algorithm begins from the goals,
but because of the SLM structure it does not need to calculate the previous states. In
the step of locating the actions that satisfy the current goals, all the services that have
a positive value in the relevance function are considered as candidates. The best
service is chosen based on the QoS characteristics. The process continues until it
reaches the initial state.
Another approach described in [17] uses model checking techniques for producing
the plan. The algorithm consists of four steps. In the first step, the goal and the initial
states are defined. In the second step, the model of the process on which the checks
will be running is extracted. The web services that could be used for the domain are
automatically detected and the state space where the solution is searched is
constructed. Information on the services is retrieved by the ontologies and is inserted
to the model. In the third step, the search algorithm in the plan space is executed and
