Application of MBSE to model Hierarchical AI Planning problems in HDDL

The recent improvements of hierarchical AI planning open the path to new and exciting application in different areas of expertise. One domain with daring and complex planning and scheduling problems is the definition of operations for space exploration systems. For this specific application, the Hierarchical Definition Domain Language (HDDL) may be the most suitable AI planning language to be adopted. However , the design and writing of problems and domain files for HDDL is a complex task. They require a skilful designer to write and check the consistency of the syntax. Moreover, sharing and modifying HDDL files can be a complicated task, and it may lack traceability of the modifications, making the overall process prone to errors. On the other hand, planning languages like HDDL and PDDL are hardly ever studied in the university curricula by most space systems engineers, the architects of the concepts of operations of space systems. The work proposed in this paper contributes to filling the gap between space operations engineers and the AI planning potentialities to solve planning and scheduling problems applied to space exploration systems. The problem and domain files typical of HDDL are built up from the formalism of SysML, a general-purpose architecture modelling language for System Engineering. SysML is effectively used as modelling language in Model-Based System Engineering (MBSE) to study and design the mission architecture of a space mission. The methodology presented is applied to an analogue space robotic mission, where a collaborative drone and a rover need to explore an unknown environment. The final aim of the method is to transfer the "human knowledge" in the planning problem and showing the capabilities of MBSE applied to Knowledge Engineering (KE) of AI planning problems.