title: Translating RDDL to JANI
keywords: stochastic planning, stochastic model checking, dynamic Bayesian networks, translation
topics: Case studies and Applications , Dependability, security and performance , Languages
committee: Ernst Moritz Hahn


The Relational Dynamic Influence Diagram Language (RDDL)
is a modelling language which is has found widespread use in the domain of stochastic planning. On the other hand, the modelling language JANI
has started to become a standard exchange format for models in the area of stochastic model checking. Obviously, there are overlaps in these two areas. In particular, it would be useful to be able to make the many models which exist in the area of planning and which are modelled in RDDL available for the application of model checking methods. To do so, it would be useful to translate RDDL to JANI. However, doing so is a nontrivial tasks, because both languages are conceptually different. Therefore, while a transformation from the predecessor language of RDDL, PDDL
to JANI exists, there is at present no such transformation from RDDL TO JANI.

The purpose of this project is to implement a translation from RDDL to JANI. Depending on the concrete objectives decided before starting the project and the progress achived, it could consist of:
- deciding which relevant subset of RDDL can be transformed to JANI, depending on which features are used by a large variety of existing models and which constructs are suitable for translation,
- implementation of the translation,
- proposing an extension to JANI in case this would be useful for the faithful translation of certain RDDL features which could presently not be translated to JANI or only be translated in a way which is undesirable,
- in case an extension of JANI is proposed, a discussion about how an efficient exploration of the state space of such extended JANI models could be implemented, possibly together with a proof-of-concept implementation,
- choice of test cases,
- evaluation of the approach in terms of resources required for the transformation and the quality of the resulting JANI files in terms of e.g. size and readibility.