Abstract: In this paper, based on a new smoothing function of the well-known nonsmooth vector-valued min-function, a non-interior-point continuous algorithm for second-order cone programming is presented. The features of this method are as follows: firstly, the starting point can be chosen arbitrarily; secondly, at each iteration, only one system of linear equations is performed for searching an improving direction; finally, global, strong and superlinear convergence are obtained without assumption of strict complementarity. The numerical results demonstrate the effectiveness of the algorithm.

Key words: second-order cone programming, continuous algorithm, vector-valued min-function, superlinear convergence