OPTI global optimisation, multi-objective optimisation
global optimisation, multi-objective optimisation
The OPTI team resulted from the grouping of the MEO and ROOM teams in September 2010.
The OPTI team’s objective is to develop knowledge designed to resolve complex optimisation problems effectively. It is directed towards a field of problems ranging from the processing of difficult large-scale NP problems to the certification of computations, with the common theme of developing effective methods, algorithms and software for resolving underlying optimisation problems.
Following in the footsteps of the recognition and successful results obtained by MEO and ROOM, the OPTI team is pursuing avenues of research to lay the foundation stones that they wish to build upon.
Global optimisation. This area concerns certified scientific computation based on interval analysis and global optimisation for digital non-linear, non-convex problems. The chosen approach consists in developing methods of constraint satisfaction and interval analysis and creating hybrid solutions using conventional optimisation tools. The challenge in the next four-year period is to demonstrate that certified computation tools are capable of resolving optimisation problems under difficult constraints.
Multi-objective optimisation. Arising from combinatorics problems with several optimisation objectives, the research carried out in this area characterises ROOM’s main scientific activities. On the strength of its recent productions on exact and meta-heuristic methods dedicated to this problem framework, the team has initiated work on approximate algorithms with performance guarantees, the characterisation of digital instances and the taking into account of preferences. These aims are positioned within the objectives of the ANR Blanc Guepard project (2009-2013), which govern our work in the four-year period 2012-2015.
Software. The development and distribution of effective software will be the team’s main concerns. This activity is MEO’s stock in trade and addresses the gaol interval arithmetic library and the digital constraint solving library realpaver. One ambition is to advance these generic systems to perform parallel computations and tackle optimisation problems. With a more specific vision, ROOM is incorporating a range of operational tools into the recife system to study the capacity of railway infrastructures. The application which is a multi-criteria decision-making aid system that may make advances by in particular integrating productions proposed by the team to resolve this problem.
Applications. Experiments are being carried out on real problems to validate the research work and also to extract new problems. The fields concerned constitute opportunities which have arisen and in which investment has been granted with a view to acquiring the necessary knowledge. These systems are often complex, such as railway transport, communication networks, production, human-machine systems, design, automatic systems and robotics. This area will be pursued as a priority in this variety of fields but will not necessarily put clued research into openings such as for example hybrid systems in biology.
The OPTI team also wishes to embark upon work which lies at the crossroads between different areas of knowledge. It would like to tackle mixed variable problems (continuous and discrete) and to develop hybrid resolution methods (local and global, linear and non-linear). One objective is to share a common culture by 2012 and to start work on precisely identified points.
Finally, OPTI has set itself the new goal of developing research work on high-performance calculations for optimisation problems. Results have already been obtained in parallel algorithmics for solving digital problems.