STOCHASTIC PROGRAMMING

Description

With each success story, optimal design and control methodologies have enjoyed increased use in various industries such as aerospace, automotive, engineering, environment, manufacturing, health care, transportation, travel, and weapons.

In most of these industrial applications, however, it is assumed that all information needed to formulate and solve a design or control problem is deterministic, that is, all information is known. The final solution is expected to be not only optimal, but also reliable.

In reality, uncertainty exists everywhere. Stochastic programming problems arise from applications with inherent uncertainty. In experiments, we may not know all the design parameters. In simulations, we may not know or have perfect descriptions of the input parameters to computer-based models. Likewise, it is possible to have uncertainty within the computer-based models themselves.

By formulating optimal design and control problems so that uncertain information is reconciled, it may be possible to generate optimal solutions that are robust and reliable to some safety margin. Our goal is not only to formulate stochastic programming problems that intelligently "incorporate" uncertain information, but to develop robust and efficient stochastic programming methods that solve these problems.

Project Members

Students

Chris Siefert, University of Illinois at Urbana-Champaign (Summer Intern 2001)

Related Projects

EUVL