A Dual Decomposition Algorithm for Separable Nonconvex Optimization Using the Penalty Function Framework

We propose a dual decomposition method for solving separable nonconvex optimization problems that arise e.g. in distributed model predictive control over networks. We first derive a new sequential convex programming (SCP) scheme based on penalty function approach to handle nonconvexity. Then, we combine this SCP scheme with a dual decomposition algorithm to obtain a two-level decomposition algorithm. The global convergence of this algorithm is analyzed under standard assumptions. Some preliminary numerical results are also given to illustrate the theoretical results.