Controllability in Partial and Uncertain Environments

Controller synthesis is a well studied problem that attempts to automatically generate an operational behaviour model of the system-to-be that satisfies a given goal when deployed in a given domain model that behaves according to specified assumptions. A limitation of many controller synthesis techniques is that they require complete descriptions of the problem domain. This is limiting in the context of modern incremental development processes when a fully described problem domain is unavailable, undesirable or uneconomical. Previous work on Modal Transition Systems (MTS) control problems exists, however it is restricted to deterministic MTSs and deterministic Labelled Transition Systems (LTS) implementations. In this paper we study the Modal Transition System Control Problem in its full generality, allowing for nondeterministic MTSs modelling the environment´s behaviour and nondeterministic LTS implementations. Given an nondeterministic MTS we ask if all, none or some of the nondeterministic LTSs it describes admit an LTS controller that guarantees a given property. We show a technique that solves effectively the MTS realisability problem and it can be, in some cases, reduced to deterministic control problems. In all cases the MTS realisability problem is in same complexity class as the corresponding LTS problem.