A globally and superlinearly convergent feasible QP-free method for nonlinear programming (Q2572703)

The author proposes a quadratic programming (QP)-free type algorithm which solves the problem of minimizing a smooth function subject to smooth inequality constraints. In constrast successive quadratic programming methods, each iteration of this algorithm only needs to solve systems of linear equations which are derived from the equality part in the Karush-Kuhn-Tucker first order optimality conditions. If the quasi-Newton direction is zero, a descent direction can be obtained by dropping a constraint from the active set at the current iterate. A high order correction direction is computed by solving a system of linear equations with small scale to assure superlinear convergence of the algorithm and to prevent Maratos effect. Unlike in the works of \textit{J. N. Herskovits} [Math. Program. 36, 19--38 (1986; Zbl 0623.90070)] and \textit{E. R. Panier, A. L. Tits} and \textit{J. N. Herskovits} [SIAM J. Contr. Optim. 26, 788--811 (1988; Zbl 0651.90072)], the initial point is not necessary to be an interior point, and the linear search replaces the arc search in the paper of Panier, et al. [loc. cit.]. Global convergence is proven under some suitable conditions, too. Some numerical results based on the algorithm are given. They show that the algorithm is effective.