TR2015-060
Decomposition via ADMM for Scenario-Based Model Predictive Control
-
- "Decomposition via ADMM for Scenario-Based Model Predictive Control", American Control Conference (ACC), DOI: 10.1109/ACC.2015.7170904, July 2015, pp. 1246-1251.BibTeX TR2015-060 PDF
- @inproceedings{Kang2015jul,
- author = {Kang, J. and Raghunathan, A.U. and {Di Cairano}, S.},
- title = {Decomposition via ADMM for Scenario-Based Model Predictive Control},
- booktitle = {American Control Conference (ACC)},
- year = 2015,
- pages = {1246--1251},
- month = jul,
- publisher = {IEEE},
- doi = {10.1109/ACC.2015.7170904},
- isbn = {978-1-4799-8685-9},
- url = {https://www.merl.com/publications/TR2015-060}
- }
,
- "Decomposition via ADMM for Scenario-Based Model Predictive Control", American Control Conference (ACC), DOI: 10.1109/ACC.2015.7170904, July 2015, pp. 1246-1251.
-
MERL Contacts:
-
Research Area:
Abstract:
We present a scenario-decomposition based Alternating Direction Method of Multipliers (ADMM) algorithm for the efficient solution of scenario-based Model Predictive Control (MPC) problems which arise for instance in the control of stochastic systems. We duplicate the variables involved in the non-anticipativity constraints which allows to develop an ADMM algorithm in which the computations scale linearly in the number of scenarios. Further, the decomposition allows for using different values of the ADMM stepsize parameter for each scenario. We provide convergence analysis and derivethe optimal selection of the parameter for each scenario. The proposed approach outperforms the non-decomposed ADMM approach and compares favorably with Gurobi, a commercial QP solver, on a number of MPC problems derived from stopping control of a transportation system.
Related News & Events
-
NEWS MERL researchers present 10 papers at the American Controls Conference Date: July 3, 2015
MERL Contacts: Daniel N. Nikovski; Yebin Wang; Stefano Di Cairano; Arvind Raghunathan; Avishai WeissBrief- MERL researchers presented 10 papers at the American Controls Conference, in Chicago, USA. The ACC is one of the most important conferences on control systems in the world. Topics ranged from theoretical, including new algorithms for Model Predictive Control and Co-Design, to applications including spacecraft control and HVAC systems.