Service Composition with PDDL Representations and Visualization over Videogame Engines

We devise a succinct knowledge representation framework based on a nondeterministic variant of a well-known Artificial Intelligence formalism, called PDDL, for representing dynamic domains in Planning. We represent the environment and the agents' (high-level) behavior as distinct PDDL action domains and exploit service composition techniques, to compose agent behaviors so as to realize a collective behavior of interest to the user. Specifically, we characterize the computational complexity of the problem and give effective algorithms for solving it, taking advantage of the succinct representation in PDDL. We explore the visualization of service composition over this framework based on the videogame metaphor of virtual worlds using a popular engine. The execution then of agents as game characters provides a form of procedural attachment of atomic actions to concrete interactions within a realistic 3D space.