Attainable solutions and model sensitivity

The system-analytical approach using so-called attainability domains helps researchers to oversee how sensitive a model’s objective function is to selecting a particular control strategy; in particular, it allows sub-optimal (i.e., close-to-optimal) solutions to be constructed and analyzed. ASA research deals both with theoretical foundations of construction of attainability domains and applications of this approach to economy-climate models.

In [1] the problem of constructing the attainability domain of a stochastic system in the absence of variable continuity and under asymptotic constraints was considered; researchers obtained a constructive description of the attainability domain’s structure and proved its robustness with respect to relaxation of constraints.

Boundary conditions are sometimes uncertain in human-Earth systems models. ASA researchers are developing methods of constructing guaranteed controls under uncertainty and incomplete information. In 2014 first results on a new approach of guaranteed feedback control of dynamic systems were published by [2] and [3], which suggested a constructive algorithm for linear control dynamic systems.

Sensitivity of the solution toward parameters of the objective function in the problem of maximizing a time-averaged objective function was studied by [4] and [5]. An optimal solution was selected from among stationary strategies, and a special type of periodic cyclic trajectory, so-called level cycles, were derived; and phase transitions between two optimal strategies were revealed.

Last edited: 12 March 2015