Fast fasible direction methods, with engineering applications

Prof. Andre L. Tits

Department of Electrical Engineering and Institute for Systems Research, University of Maryland, College Park

Optimization problems arising in engineering applications often present distinctive features that are not exploited, or not accounted for, in standard numerical optimization algorithms and software codes. This paper focuses on a few of these: multiple objectives, functional constraints or large number of ``related'' constraints, and requirement that all iterates be ``feasible''. A basic scheme for efficiently tackling inequality constrained optimization while forcing feasible iterates is discussed and various extensions are proposed.