Abstract
We recall some of the multi-criteria and multidisciplinary optimization formulations with emphasis on algorithmic and implementation issues. We use a partial controllability result which states that the vicinity of any point in the optimization space can be reached in finite time from any other point in the space following a trajectory solution of a particular second order dynamical system. We use this continuous point of view to propose a class of heuristic optimization algorithms also addressing global search issues. For multi-criteria problems this means that any point on a Pareto front can be reached even if the front is non-convex as far as the optimization algorithm has inertial features. The paper also recalls the conditions under which a Pareto front can be accessed by descent methods applied to multicriteria problems using the weighted sum approach. The question of the density of the distribution of the points over the front is also addressed. The approach is illustrated on academic examples as well as on a multi-criteria distributed inverse problem for a heterogeneous data fusion problem.