The work presented in this thesis deals with modelling and decision support issues in the context of complex systems. Specifically, we investigate the coupling of agent-based simulations and inverse problem resolution methods. Thus, a generic architecture of decision support system is proposed. We identify two problems related to the implementation of this architecture : the validation of simulation parameters and the observation of agent-based simulations. We offer some answers to solve them. The notion of inverse problem is redefined in the context of decision support in a multi-model framework. Therefore, a resolution method and heuristics, based on ideas from works on abductive reasoning, are proposed. This work is then applied to the conception of a decision support system dedicated to forensic entomology. This research area, midway between forensic sciences and entomology, aims to develop post-mortem estimation methods based on entomological indices, i.e., insects (mainly necrophagous diptera) or insect remains, sampled on or around the cadaver. In the context of this work, done in collaboration with forensic entomology laboratory of the Legal Medicine Institute of Lille, we present : * an agent-based model of necrophagous diptera in a complex environment such as a dead body in a specific ecosystem, taking into account more parameters than conventional methods, * an implementation of the decision support system architecture mentioned above, based on this model.
