1: \begin{abstract}

2: %Outline:

3: %\begin{itemize}

4: %\item Nonlinearly constrained optimization: importance.

5: %\item Unified framework allows us to express a variaty of solvers.

6: %\item Framework implications for software design; simplify implementation of new algorithms.

7: %\item Example: new look at funnel; methods.

8: %\item Convergence analysis of funnel methods and similarity to filter methods motivated by framework.

9: %\item Numerical results.

10: %\end{itemize}

11: We consider nonlinearly constrained optimization problems and discuss a generic double-loop framework consisting of four algorithmic ingredients that unifies a broad range of nonlinear optimization solvers. This framework has been implemented in the open-source solver \texttt{Uno}, a Swiss Army knife-like C++ optimization framework that unifies many nonlinearly constrained nonconvex optimization solvers.

12: We illustrate the framework with a sequential quadratic programming (SQP) algorithm that maintains an acceptable upper bound on the constraint violation, called a funnel, that is monotonically decreased to control the feasibility of the iterates. Infeasible quadratic subproblems are handled by a feasibility restoration strategy. Globalization is controlled by a line search or a trust-region method.

13: We prove global convergence of the trust-region funnel SQP method, building on known results from filter methods. We implement the algorithm in \texttt{Uno}, and we provide extensive test results for the trust-region line-search funnel SQP on small \texttt{CUTEst} instances. 

14: %\sven{Update abstract at end!}

15: \end{abstract}

16: