ID4CS
ID4CS Integrative Design for Complex Systems
Projet ANR, appel COSINUS 2009
| Dates: 2009 - 2012 |
Funding:  |
| Parterns: |
| Web site: http://www.irit.fr/ID4CS |
| Contact for SMAC: Marie-Pierre Gleizes |
Objectives of the Research Project
The ID4CS project has the ambition to propose a modelling and simulation environment for designing complex systems such as aircraft. The complexity of the task comes from the following features of the problem: multiple disciplines are involved, multiple objectives must be reached, multiple criteria are associated to the models, there are multiple levels of design and multiple optimization strategies available.
The aim of ID4CS is to propose a self-adaptive, distributed and open architecture enabling the integration of the previous multi-dimensional aspects. All the elements work locally to a collective process solving while interacting cooperatively with their neighbours’ elements, using meta-information and while learning from experiences. The originality of the project lies in enabling several different optimization techniques to collaborate within a same system and taking into account all different dimensions of the problem during the resolution.
This is done by using adaptive multi-agent technology to solve the complex system design. In order to comply with uncertainty requirements, the resolution process may switch from one granularity level to a neighboring granularity level. All along the simulation, agents of the appropriate level of granularity will be activated. The levels interact in the following way: the upper level provides strategies or some information to guide the lower level resolution process. When the designer’s request will be sent to the system, all the agents of the system will find the way to interact in order to build the right organization of models to satisfy the designer’s request. Because numerous search methods and models exist, the system will be open. This means that it will be designed with only some methods and other could be incrementally added into the system. Moreover, the adaptive capability of agents allows the designers to interact dynamically with the system during its resolution process in order to adjust the problem specifications or to observe the current solution state. The designer co-constructs the solution of his problem design with the system and for facilitating the interaction GUIs and a management tool will be developed. This workbench will be evaluated on the preliminary aircraft design domain.
We intend to address together, within a single framework, five important dimensions of today’s real-world optimization problems: multi-level, multi-disciplinary, multi-search-method, multi-objective, and multi-attribute.