Risk modeling and UNcertainty analysis using Evidential networks and Semantic Knowledge representations

Principal investigators: Mohamed Sallak and Domitile Lourdeaux (Heudiasyc)




The RUNESK project aims at proposing a combined approach using a semantic model and a quantitative risk model for risk assessment of Systems-of-Systems based on imprecise risk parameters. More specifically, it proposes a model to analyze the risk in two layers, a layer for semantic model and a layer based on Valuation Based Systems (VBS) to perform the computations.


The proposed combined risk analysis approach provides important information such that occurrence probability of all events, critical paths, ranks of barriers (technical barriers, human barriers, etc). By knowing the importance of each barrier we can decide which one should have the highest priority. Moreover, knowing the best place to add a barrier is very helpful to reduce the probability of occurrence of some specific or critical events.

  • Regarding the combined risk assessment approach, M. Sallak and D. Lourdeaux have supervised two Master theses: Edward Gregorius and Francesco Inamoreti, whose work [Edward Gregorius, ESREL 2016] focuses on developing such a combined model.
  • Regarding the use of graphical models in risk assessment and planning: M. Sallak and D. Lourdeaux are currently supervising Rémi Lacaze, whose PhD work focuses on combining planning methods and graphical probabilistic models (such as Bayesian networks and VBS) for identifying critical situations in virtual environments under uncertainty.


We are currently pursuing the latter direction and are trying to obtain a generic risk assessment method for handling critical situations in virtual environments under uncertainty.