Laboratoire de Génie Informatique et d’Automatique de l’Artois

Ph.D. thesis of Alexandre VEREMME

Interests and uses of the theory of belief functions for decision support systems based on multi-agent systems. Application to forensic entomology.

Starting date: 1 October 2007
Keywords: TODO

This work is based on the agent simulations (ABS), the cellular automata based simulations (CABS) and the theory of belief functions. The purposes consist in proposing solutions to complex problems and developping a general architecture for decision support system (DSS) organized around these three tools. Because this system resorts to the theory of belief functions (also called the evidence theory), we called it EDSS, i.e., evidential decision support system. The EDSS tries to explain an observation made on a complex real system. It uses:- ABS and CABS to simulate the complex system of study;- the belief function theory to represent and manipulate information within and around the simulations.The architecture is validated through a decision support system, named ForenSeek, dedicated to experts in forensic entomology. Forensic entomology is a method for estimating the time of death from determining the ages and species of insects collected on a cadaver. This technique faces with a set of complex phenomena, not easily discernible by the experts and having an impact on their final decisions. The EDSS ForenSeek uses the ABS and CABS to simulate the corpse decomposition. It incorporates a large number of ecosystem parameters and tests a large quantity of biological models. The theory of belief functions is central to this information fusion system. First results from real forensic expert’s reports are presented.

Involved research axes:


Defense took place the 08/12/2010 pm31 13:30 Prestige room - FSA - Béthune


  • Rapporteur Arnaud MARTIN IRISA, Université de Rennes 1
  • Rapporteur Jean-Pierre MÜLLER CIRAD de Montpellier
  • Examinateur Zied ELOUEDI Institut Supérieur de Gestion de Tunis
  • Invité Philippe KUBIAK LAGIS, Université Lille 1
  • Invité Benoît VINCENT Institut de Recherche Criminelle de la Gendarmerie Nationale
  • Advisor Daniel JOLLY LGI2A, Université d'Artois
  • Co-supervisor Daniel DUPONT HEI, Université Catholique de Lille
  • Co-supervisor Eric LEFEVRE LGI2A, Université d'Artois