SBIR-STTR Award

This Small Business Technology Transfer Phase I project will address the problem of creating robust and efficient software for the solution of large-scale nonlinear optimization problems. Taking advantage of recent advances in algorithm design, the project will investigate novel versions of both interior-point and active-set methods for nonlinear optimization, and will examine an innovative integrated approach that takes advantage of both methods to achieve greater speed and reliability than are available from current single-method codes. Optimization software resulting from this research is expected to have commercial applications to difficult nonlinear problems in such areas as network planning, optimal power flow, computer-aided design, and aerospace engineering, as well as in applications for decision analysis in such areas as finance and revenue management. Broader impacts of the activity include enhanced understanding of optimization techniques that play a key role in engineering and commerce; an advance for NSF's educational goals by providing support to a postdoctoral researcher; and commercial software products whose further development can be supported by sales revenues

This Small Business Technology Transfer Phase II research project will address the design and creation of integrated nonlinear optimization software that combines complementary approaches to nonlinear optimization to achieve robust performance over a wide range of application requirements. The work will concentrate on the area of smooth nonlinearly constrained optimization, which arises directly in numerous applications and as a sub-problem in mixed-integer nonlinear programming and global optimization. The work will employ both mathematical convergence analyses and extensive testing on problems of practical interest. Results of the research will take nonlinear optimization software to a new level, based on an adaptive and versatile collection of algorithms in contrast to the single-algorithm approaches employed by current optimization packages.Nonlinear optimization models arise in diverse areas of science such as medical imaging, oceanography, crystallography, and climate modeling, and in almost all areas of engineering, chip feature placement for semiconductor manufacturers to energy management for electric and gas utilities. Nonlinear optimization is also rapidly becoming a key tool in decision analysis in such areas as finance and revenue management. By enabling optimization packages to be more fexible and more reliable, this research will lead to stronger support for current nonlinear optimization applications while making new, more ambitious applications possible.