Fully Distributed Algorithms for Densely Coupled Optimization Problems in Sparse Optimization and Transportation Applications

Author/Creator ORCID

Date

2021-01-01

Department

Mathematics and Statistics

Program

Mathematics, Applied

Citation of Original Publication

Rights

This item may be protected under Title 17 of the U.S. Copyright Law. It is made available by UMBC for non-commercial research and education. For permission to publish or reproduce, please see http://aok.lib.umbc.edu/specoll/repro.php or contact Special Collections at speccoll(at)umbc.edu
Distribution Rights granted to UMBC by the author.
Access limited to the UMBC community. Item may possibly be obtained via Interlibrary Loan thorugh a local library, pending author/copyright holder's permission.

Subjects

Abstract

Distributed algorithms are gaining increasing attention with broad applications indifferent areas such as multi-agent network systems, big data, machine learning, and distributed control systems, among others. Most of the distributed optimization algorithms developed assume a separable structure for the underlying optimization problems, and certain coupled optimization problems are often solved via partially distributed schemes. In this thesis, we develop fully distributed algorithms for densely coupled optimization problems in two topics, namely, column partition based sparse optimization problems and transportation applications. Firstly, we develop two-stage, fully distributed algorithms for coupled sparse optimization problems including LASSO, BPDN and their extensions. The proposed algorithms are column partition based and rely on the solution properties, exact regularization, and dual formulation of the problems. The overall convergence of two-stage schemes is shown. Numerical tests demonstrate the effectiveness of the proposed schemes. Secondly, we develop fully distributed algorithms for model predictive control (MPC) based connected and autonomous vehicle (CAV) platooning control under linear and nonlinear vehicle dynamics. In the context of linear vehicle dynamics, the underlying optimization problem of the MPC is a densely coupled, convex quadratically constrained quadratic program (QCQP). A decomposition technique is developed to formulate the densely coupled QCQP as a locally coupled convex optimization problem. We then develop operator splitting method based schemes to solve this problem in a fully distributed manner. Particularly, to meet challenging real-time implementation requirements, a generalized Douglas-Rachford splitting method based distributed algorithm is proposed, along with initial state warm up techniques. Under nonlinear vehicle dynamics, the underlying problem is a densely coupled, nonconvex optimization problem. We develop a sequential convex programming based fully distributed optimization algorithms. Control and closed loop stability analysis are carried out for both linear and nonlinear vehicle dynamics. Numerical tests performed for possibly heterogeneous CAV platoons demonstrate the effectiveness of the proposed schemes.