Handling infeasibility in a large-scale nonlinear optimization algorithm (Q430999)

In constrained optimization, one aims to find the lowest possible value of an objective function within a given domain. Practical nonlinear programming algorithms may converge to infeasible points, even when feasible points exist. Therefore, optimization users that wish to find feasible and optimal solutions of practical problems usually change the initial approximation and/or the algorithmic parameters of the algorithm when an almost infeasible point is found. The effectiveness of this trial-and-error process is, in part, related with the ability of the algorithm of stopping quickly when the generated sequence is fadded to converge to an infeasible point. It is sensible to detect this situation as quickly as possible, in order to have time to change initial approximations and parameters, with the aim of obtaining convergence to acceptable solutions in further runs. In this paper, a recently introduced augmented Lagrangian algorithm is modified in such a way that the probability of quick detection of asymptotic infeasibility is enhanced. The modified algorithm preserves the property of convergence to stationary points of the sum of squares of infeasibilities without harming the convergence to Karush-Kuhn-Tucker points in feasible cases.