Diagnosis in fault tolerant and autonomous architectures
Diagnosis cannot be seen as an isolated task but an ingredient into fault management architectures. It takes part in the solutions produced for tasks such as on-board recovery, condition monitoring, maintenance and prognosis, repair and therapy planning. The contribution of diagnosis in such architectures means close links with decision tasks such as control and planning and calls for innovative integrations that are among the DISCO team interests.
The analysis of the reconfigurability of a controlled system, possibly in real time, when a fault occurs is at the core of our work. The aim is to determine new trajectories and the appropriate control laws to guarantee the stability of the system and to keep the system close to the desired state. Hierarchical fault tolerance is also investigated, so that a system can be reconfigured at the functional level.
A second aspect, at the core of autonomy, is the articulation of diagnosis and planning as decision making. Planning generates a control program, also known as a plan, which describes the sequence of actions necessary to achieve some predeﬁned goals. Diagnosis offers the capability to detect, isolate and sometimes identify defects or more generally the root causes of one or more discrepancies, in models, plans or reasoning inferences. In the last two decades, most architectures for autonomy have derived from an adhoc and empirical partition of the agent and its world in their representations and operations. This partitioning leads to a set of modules, typically a planner/scheduler, a health monitoring system, and an executive.
The DISCO team aims at justifying the use of diagnosis when it is needed, and at articulating diagnosis with other functional modules within a generic architecture for autonomy. This represents a cornerstone to achieve the truly robust autonomy of tomorrow.