Direct iterative solver for model predictive control
Abstract
Abstract
Model Predictive Control (MPC) has begun to receive significant attention as a tool in real-time solver applications, as the betterment of technology and algorithms have allowed much faster operation time. This work explores the use of Lemke’s Algorithm, with an augmented p-vector, and q-based block-partitioning to solve the Linear Complementarity Problem (LCP) that is derived from the condensed MPC formulation. The conversion of MPC to LPC is shown in the “Methodology” section. The iteration count and comparison to Matlab ‘quadprog’ is displayed. It is shown that iteration count is less than or equal to the rank(H), and the proposed algorithm always matched quadprog.
Description
Department of Electrical & Computer Engineering
Rights
File access restricted due to FERPA regulations