Part-Whole Hierarchical Modularization of Fault-Tolerant and Goal-Based Autonomic Systems

The paper examines current trends in autonomic space software systems and proposes the adoption of a hierarchical state based formalism which allows the different paradigms employed in the field to meet seamlessly. It is not clear, for example, how goal- based on-board autonomy, originally conceived for deductive-flavored systems which may not be ground controlled for indefinite portions of time, be able to combine with model based engineering, which best fits current industrial design strategies. Other aspects, such as closed loop discrete control and fault tolerance do not easily lend themselves to modularity. The paper shows that, by employing the proposed formalism, goals can be decomposed and distributed in a very natural way among different modules. Each module is, at the same time, both a controller and a controllable part of the whole system, allowing to partition the closed loop control flow at different levels of complexity. Formal verification is also possible by employing goals as state based constraint in the implementation phase.