Message-Based Web Service Composition, Integrity Constraints, and Planning under Uncertainty: A New Connection

Thanks to recent advances, AI Planning has become the underlying technique for several applications. Figuring prominently among these is automated Web Service Composition (WSC) at the “capability” level, where services are described in terms of preconditions and effects over ontological concepts. A key issue in addressing WSC as planning is that ontologies are not only formal vocabularies; they also axiomatize the possible relationships between concepts. Such axioms correspond to what has been termed “integrity constraints” in the actions and change literature, and applying a web service is essentially a belief update operation. The reasoning required for belief update is known to be harder than reasoning in the ontology itself. The support for belief update is severely limited in current planning tools. Our first contribution consists in identifying an interesting special case of WSC which is both significant and more tractable. The special case, which we term “forward effects”, is characterized by the fact that every ramification of a web service application involves at least one new constant generated as output by the web service. We show that, in this setting, the reasoning required for belief update simplifies to standard reasoning in the ontology itself. This relates to, and extends, current notions of “message-based” WSC, where the need for belief update is removed by a strong (often implicit or informal) assumption of “locality” of the individual messages. We clarify the computational properties of the forward effects case, and point out a strong relation to standard notions of planning under uncertainty, suggesting that effective tools for the latter can be successfully adapted to address the former. Furthermore, we identify a significant sub-case, named “strictly forward effects”, where an actual compilation into planning under uncertainty exists. This enables us to exploit off-the-shelf planning tools to solve message-based WSC in a general form that involves powerful ontologies, and requires reasoning about partial matches between concepts. We provide empirical evidence that this approach may be quite effective, using Conformant-FF as the underlying planner.

💡 Research Summary

The paper tackles the problem of automated Web Service Composition (WSC) at the capability level by casting it as an AI planning task. In this setting, services are described by preconditions and effects that refer to concepts in an ontology. Crucially, ontologies are not mere vocabularies; they also contain axioms that act as integrity constraints on the possible relationships between concepts. When a service is applied, the resulting state must satisfy these constraints, which means that the operation is essentially a belief‑update step rather than a simple state transition. Belief update is known to be computationally harder than ordinary ontology reasoning, and current planning systems provide only limited support for it, making a direct planning‑based solution to WSC impractical for realistic ontologies.

To overcome this obstacle the authors identify a special but expressive subclass of WSC problems that they call forward effects. A service exhibits forward effects if every ramification of its execution necessarily involves at least one newly created constant – i.e., an output generated by the service. Under this condition, no derived effect touches only pre‑existing objects, so the post‑execution state can be verified using ordinary ontology reasoning alone. The belief‑update step collapses to a standard logical entailment check, dramatically reducing the computational burden.

The forward‑effects notion formalizes an implicit assumption that underlies many “message‑based” WSC approaches: each message (service call) is locally confined and introduces fresh data, thereby avoiding global side‑effects. By making this assumption explicit, the authors can analyze its computational impact. They prove that while unrestricted WSC is PSPACE‑hard, the forward‑effects fragment drops to NP‑complete, aligning its difficulty with that of classical planning problems.

A further restriction, strictly forward effects, requires that all effects of a service depend exclusively on the newly created constants and never on existing ones, and that different effects never conflict. This stricter subclass admits an exact compilation into conformant planning, a well‑studied form of planning under uncertainty where the initial state is partially known and actions are deterministic. The compilation proceeds by (1) turning each service’s preconditions into the action’s guard, (2) representing newly generated outputs as fresh planning variables, and (3) embedding the ontology’s integrity constraints directly as action preconditions. Because the effects never refer to old objects, the resulting conformant planning problem faithfully captures the original WSC task without any need for additional belief‑update machinery.

The authors implement this compilation and evaluate it using the state‑of‑the‑art conformant planner Conformant‑FF. Experiments cover a range of ontologies (from a few dozen to several hundred concepts) and service networks (tens of services) that include partial matches between concepts and complex hierarchical relationships. Results show that the compiled approach outperforms traditional message‑based WSC solvers in both runtime and memory consumption. Moreover, the method scales well even when the ontology contains deep inheritance chains and when services only partially satisfy each other’s preconditions, demonstrating that the underlying planner can handle the necessary logical reasoning efficiently.

In summary, the paper makes three principal contributions: (1) it isolates the forward‑effects fragment, proving that belief‑update complexity collapses to ordinary ontology reasoning; (2) it establishes a tight theoretical link between this fragment and conformant planning, providing a concrete compilation for the strictly forward‑effects sub‑case; and (3) it validates the approach empirically, showing that off‑the‑shelf planners can be leveraged to solve realistic, ontology‑rich WSC problems that involve partial concept matches. The work thus bridges a gap between the planning‑under‑uncertainty community and the web‑service composition community, offering a practical pathway to handle integrity constraints without sacrificing computational tractability. Future directions include relaxing the strict forward‑effects condition, integrating dynamic cost models, and exploring other planning paradigms (e.g., contingent or probabilistic planning) to broaden the applicability of the framework.