COIN-OR Open-Source Project

Mixed-Integer Nonlinear Optimization Toolkit

Minotaur is an open-source C++ software framework and solver toolkit for mixed-integer nonlinear programming (MINLP) problems, providing parallel algorithms, relaxations, and presolving routines.

Download & Install User Guide Publications
Minotaur Solver Logo

Implemented Solvers

mbnb — Convex MINLP Solver

Nonlinear programming-based branch-and-bound solver designed for convex mixed-integer nonlinear problems.

mqg — Branch-and-Cut Engine

An LP/NLP-based Quesada-Grossmann branch-and-cut algorithm for convex MINLPs that generates cutting planes on the fly.

mglob — Global Optimizer

Spatial branch-and-bound solver targeting global optimization of Mixed-Integer Quadratically Constrained Quadratic Programs (MIQCQP).

mmultistart — Multistart Heuristic

A heuristic search algorithm that runs local optimization searches from multiple start points for general non-convex MINLPs.

Toolkit Capabilities

Algorithmic Framework

Standard C++ implementations of branch-and-bound, branch-and-cut, outer-approximation, and multistart heuristics for convex and nonconvex optimization.

Automatic Differentiation

Supports nonlinear functions using automatic differentiation to evaluate first and second-order derivatives for solvers and relaxations.

Flexible Integration

Can be run as a standalone executable, called as a C++ library, or integrated with algebraic modeling languages like AMPL and Pyomo.