Multi-Agent & Multi-Objective Optimization
The team is interested in the cooperative, decentralized and distributed aspects of decisions, related to the presence of several decision centers that interact in a number of applications. The team conducts research in multi-objective mathematical programming. Within multi-agent optimisation problems, the team is also exploring the search for equilibrium solutions within the meaning of game theory that are also non-Pareto dominated. Finally, the team is interested in distributed combinatorial optimization, especially for reasons of security or respect of private data. Interdisciplinary researches on human factors in combinatorial optimization have also been carried out.
Lower and uper bounds for a biobjective integer program
Multi-objective combinatorial optimization
The single-objective case is here extended to multiple objectives. Considering the Pareto dominance, the team mainly focuses on a posteriori methods which aim at generating the whole set of non-dominated points in the objective space and a solution for each of them. The ROC team aims to design novel multi-objective mathematical programming approaches (branch-and-cut, column generation) and metaheuristics.
Some publications on multi-objective combinatorial optimization
Multi-agent project scheduling problem
Multi-agent combinatorial optimization
In several real-life contexts, there is not a single decision maker responsible for solving the whole problem. This becomes especially true as the size and the complexity of the problems solvable through combinatorial optimization techniques is larger and larger. Multi-agent combinatorial optimization aims at considering the distributed aspects of decision in a combinatorial optimization problem. The team focuses especially on the link with game theory and has worked on the charactérization of stable solutions maximizing/minimizing an objective function for scheduling or transportation problems. Another study concerns privacy preserving in distributed combinatorial optimization problems. Hybrid methods integrating both secure multi-party computation technics and shortest path technics were proposed to solve some ridesharing problems.