Some theoretical properties of an augmented lagrangian merit function

摘要: Sequential quadratic programming (SQP) methods for nonlinearly constrained optimization typically use of a merit function to enforce convergence from an arbitrary starting point. We define smooth augmented Lagrangian in which the Lagrange multiplier estimate is treated as separate variable, and inequality constraints are handled by means non-negative slack variables that included linesearch. Global proved SQP algorithm uses function. It also prove, steps unity accepted neighborhood solution when this used suitable superlinearly convergent algorithm. Finally, selection numerical results presented illustrate performance associated method. 36 refs., 1 tab.