Acronym: Virtual Reality for Safe Seveso Subcontractors

Objective: to propose computational models for simulating hazardous work situations in maintenance activities carried out by external contractors on Seveso sites. The project combines behavioural modelling of virtual humans with knowledge engineering in virtual environments.

http://www.youtube.com/watch?v=yjQ-_StpKEg

Funding: ANR/RNTL project (French National Software Technologies Network), awarded in December 2006

Duration: 4 years (starting December 2006)

Partners: UTC/Heudiasyc/UMR 6599 is the project coordinator. The other partners are INERIS, CEA-LIST, ECI (Université Paris Descartes), EbTrans, EMISSIVE, TICN, APTH and CICR.

UTC team: Domitile Lourdeaux, Dominique Lenne, Indira Thouvenin, Jean-Paul Barthès, Pavol Barger, Lydie Edward (PhD student), Kahina Amokrane, Camille Barot, Mehdi Sbaouni (research engineer), Amel Ben-Ayed (ergonomist), Morgan Fraslin (research engineer)

Students: Romain Perney, Nour El-Mawas, Pierre-Yves Gicquel, Racha Mohamad Ali, Alassane Abou Dialo, Florian Gaillot, Vincent Lanquepin, Kévin Sertier, Julien Delorme, Omar Bouslama, Camille Barot, Johan Arce


Figure: V3S virtual environment for hazardous material loading (3D developed by EMISSIVE)

Figure: V3S/FIANNA virtual environment for pipe dismantling (3D developed by Fabrice Camus and Lydie Edward)

Challenges:

• The actions of real operators are difficult to plan (virtual operators must be adaptive and autonomous)
• Difficulty in implementing research from cognitive psychology and human reliability, and cognitive models, in computational architectures (such work is lacking in France, and too cognitively oriented abroad)

Approach: We aim to account for variable and erroneous processes in normal and degraded situations. Depending on certain static or dynamic characteristics, autonomous virtual characters perceive and decide differently, and their goals change according to:

• physiological states (e.g. fatigue, stress, injuries, contamination);
• personality traits (e.g. caution, expertise).

Proposed solution:

• Modelling goal-directed activity in normal situations: the HAWAI-DL formalism
  
• Modelling activity in degraded situations: Borderline Tolerated Conditions/activities of Use [De la Garza, 2003]
  
• Error model: the CREAM method [Hollnagel, 1998]
  
• Cognition model: COCOM [Hollnagel, 1993]
• Planning
• Multi-agent system