Information Technology Reference
In-Depth Information
second step, a web service composition graph is created. The nodes of the graph are
the actions of the plan and the edges are the links that express the order constraints
among the actions. The creation of the graph is based on the information collected
from the previous step. Finally, in the last step, the composite web service functional
representation is formed using the ordering constraints that are extracted from the
composition graph. In the following paragraphs, these three steps are described in
more detail.
Fig. 1.
Converting a PDDL plan to a composite web service functional representation
The initial available information is derived from the PDDL domain and problem
files of the composition problem. For the parsing of these files, an external library,
called PDDL4J, [13] is used. The types of information that are required by the transla-
tion process are the following: a) the name of the operator, b) the parameters list, c) the
preconditions list, d) the effects list, e) the initial state and f) the goals of the problem.
Finally, the resulting plan is parsed in order to extract information concerning the ac-
tions of the plan. Exploiting the syntax of this file, information on the actions used can
easily be extracted. The data that are needed in the later steps of the algorithm involve
the timestamp of each action, which is the time step when the action will be executed
and the name, parameters and duration of it. The actions are read in the order that they
are presented in the plan, so the procedure keeps track of this order.
When all these data are retrieved, the procedure continues combining them so as
to create objects representing the steps of the plan. Every step contains the name of
the action that will be executed, the parameters with which the action is called, the
timestamp and duration of the action, the operator from which the action is derived,
the substitution imposed on the operation, the list of preconditions that must hold
for the action to be executed and the list of the effects, the facts that will change due
to the execution of the action.
The second step creates the web service composition graph. The nodes list is iden-
tical to the list of actions of the plan. In essence, the contribution of this step is the
computation of the edges, that is, the links between the actions. The general idea is to
traverse all the actions and locate cases where one precondition of an action matches
one effect of another. This ought to happen in theory because of the causal links that
