Call for papers

Diagnosis is the reasoning process that aims at determining whether
certain properties (such as the occurrence of a failure, of an
unexpected measurement, or of a deviation from prescribed behavior)
hold at a given time inside a system (fault diagnosis) and/or in its
environment (situation diagnosis) and identifying their causes.
Diagnosis reasoning is thus crucial for many reasons:
safety/maintainability in supervised systems, robustness/decision
autonomy of smart agents in their partially observed environment,
reconfiguration of business strategies after a failure... Despite the
huge spectrum of applications, Model-Based Diagnosis (MBD) problems
are generic as any diagnosis algorithm relies on a class of models, or
a modeling framework, that represents a larger class of systems. In
the last decades, many diagnostic modeling frameworks have been
proposed that are now well established.

Nowadays, the community is more and more interested in understanding
the power and the limits of such frameworks/techniques. Such studies
allow to determine in advance how the diagnosis algorithm will
behave. This is crucial especially if Diagnosis is the input of other
automated reasoning tools (like planning, scheduling,...). The
accuracy and performance of MBD algorithms depend to a large extent on
some properties of the underlying model (correctness, fidelity,
accuracy, diagnosability, predictability...). For example, the size of
the model (measured as the number of variables and constraints)
affects all major diagnostic metrics such as diagnostic accuracy,
computational performance, etc.  Tools for model analysis are
necessary to assist both in the modeling and in the computation phases
of MBD. Model analysis can be used for creating worst-case scenario
benchmarks, asserting model correctness, and facilitating automatic or
semi-automatic modeling. As such, model-based analysis can be done
with the help of a variety of AI and optimization tools such as SAT
and Max-SAT methods, search algorithms, optimal sensor placement
algorithms, and others.

The purpose of this workshop is to gather researchers from several
related Model-Based Reasoning fields such as Model-Based Diagnosis,
knowledge-compilation, satisfiability, planning and constraint
satisfaction and optimization in order to fill the gaps and exchange
ideas about which reasoning techniques are needed for the analysis of
the various models used in MBD and the implication of these analyses
to the performance of MBD algorithms.  Possible topics include but are
not limited to:

* Relevant properties of model for diagnosis:
  + structural/behavioural properties,
  + detectability, diagnosability, max-fault min-cardinality, fault
    distinguishability, finite tractability, adaptability
* Automated Model relaxation/abstractions techniques
* Model/Algorithm interfacing
* Model partitioning/decomposition for decentralized reasoning
* Performance metrics, Diagnosis accuracy
* Parametric computational complexity of MBD algorithms, anytime
  algorithms
* Meta-modeling, Meta-diagnosis
* Benchmarking of MBD algorithms, worst case scenarii synthesis