Abstract : Re-designing and assessing sustainable cropping systems has become a major challenge for agricultural researchers and farmers. In most prototyping methods, researchers are the main, or sometimes the only designers. However, more and more attempts for involving different kinds of stakeholders in participatory approaches have been accounted for. Besides, multi-criteria assessment tools are generally accepted as the solution to evaluating the overall sustainability of new cropping systems. The new method described in this thesis (DISCS for 're-Design and assessment of Innovative Sustainable Cropping Systems') implements a participatory re-designing process, while developing at the same time specific multi-criteria assessment tools for each category of stakeholders involved. Though founded on the common structure of prototyping methods, this five-step-method differs first by implementing three progress loops, at experimental field, farm then regional scale. Besides, an ongoing progress dynamic is set between research at the experimental station and crop management in farms, in order to ensure that innovative techniques actually match the farmers' expectations. Assessment at each step is performed using a scale-specific set of indicators. The DISCS method was tested on citrus production in Guadeloupe (French West Indies), to develop lower-pesticide sustainable cropping systems. Five weed management prototypes were collectively designed then tested at an experimental station based on the principles of a systematic approach. We show here that, contrary to other prototyping methods, innovative cropping systems can be the product of a farmer-researcher collaborative design. It was also verified that the innovation process benefited from involving non-agricultural stakeholders in the definition of the objectives for sustainable cropping systems. Such step by step improvement is well suited to perennial cropping systems. Using the DISCS framework should enable stakeholders to target precisely the decision-aid tools that need to be designed for a specific issue, thus ensuring a direct link between research and application. In this thesis, we present the first stages of citrus cropping system redesigning based on DISCS method; namely from the initial diagnosis, for to identify targets for improvement of the cropping system in collaboration with the stakeholders, to the multi-criteria assessment of weed management prototypes based on a set of appropriate indicators. Our work enables us to recommend sustainable weed management techniques for young orchards that correspond to the current constraints facing orchards in Guadeloupe.