Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.


Efficient Learning-Based Model Predictive Control with Affine Dynamics and Oracles using Sparse Interior Point Method


X. Zhang

Semester Thesis, HS11 (10172)

Model predictive control (MPC) provides an efficient way of solving constrained optimal control problems. In the case where the cost is quadratic, and the model and the constraints are linear, the resulting optimization problem is a quadratic program (QP-MPC). One approach applicable to time-varying systems is to solve QP-MPC online. However, it is prohibitive to use dense, general purpose solvers for large problems, as for those methods the number of operations per step in an interior point method (IPM) scales cubicly with the prediction horizon. This has given rise to the concept of sparse solvers that exploit the structure of the MPC formulation. It has been shown that the complexity to solve the Newton system arising from QP-MPC can be solved linearly in the prediction horizon by applying a primal-barrier IPM. Recent results also suggest that online methods that exploit the structure of the MPC problem can be used to control fast sampled systems. In this project, we present and implement a sparse primal-dual infeasible start IPM (PD IIPM) based on Mehrotra's predictor-corrector scheme for the learning-based MPC (LBMPC) framework. LBMPC rigorously combines MPC with statistical and learning methods, while providing guarantees on performance, safety and robustness.


Type of Publication:

(13)Semester/Bachelor Thesis

C. J. Tomlin

File Download:

Request a copy of this publication.
(Uses JavaScript)
% Autogenerated BibTeX entry
@PhdThesis { Xxx:2011:IFA_4168
Permanent link