AN EFFICIENT METHOD FOR NON-CONVEX QCQP PROBLEMS

In this paper we consider a quadratically constrained quadratic programming problem which contains convex and non-convex constraints and non-convex objective function. To obtain a global solution, we transform the original problem to a linear cone programming problem. If the solution of the new problem satisfies the constraints, it is the global optimal solution, otherwise a lower bound of the optimal value is obtained. We reduced the problem by adding a sequence of ellipsoid constraints to a family of problems whose optimal solutions converge to the global optimal solution. The convergence of the proposed method is investigated. Numerical examples are given to show the applicability of the proposed method.