A Time-certified Predictor-corrector IPM Algorithm for Box-QP

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wu, Liang, Che, Yunhong, Braatz, Richard D., Drgona, Jan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908577359200256
author Wu, Liang
Che, Yunhong
Braatz, Richard D.
Drgona, Jan
author_facet Wu, Liang
Che, Yunhong
Braatz, Richard D.
Drgona, Jan
contents Minimizing both the worst-case and average execution times of optimization algorithms is equally critical in real-time optimization-based control applications such as model predictive control (MPC). Most MPC solvers have to trade off between certified worst-case and practical average execution times. For example, our previous work [1] proposed a full-Newton path-following interior-point method (IPM) with data-independent, simple-calculated, and exact $O(\sqrt{n})$ iteration complexity, but not as efficient as the heuristic Mehrotra predictor-corrector IPM algorithm (which sacrifices global convergence). This letter proposes a new predictor-corrector IPM algorithm that preserves the same certified $O(\sqrt{n})$ iteration complexity while achieving a $5\times$ speedup over [1]. Numerical experiments and codes that validate these results are provided.
format Preprint
id arxiv_https___arxiv_org_abs_2510_04467
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Time-certified Predictor-corrector IPM Algorithm for Box-QP
Wu, Liang
Che, Yunhong
Braatz, Richard D.
Drgona, Jan
Optimization and Control
Minimizing both the worst-case and average execution times of optimization algorithms is equally critical in real-time optimization-based control applications such as model predictive control (MPC). Most MPC solvers have to trade off between certified worst-case and practical average execution times. For example, our previous work [1] proposed a full-Newton path-following interior-point method (IPM) with data-independent, simple-calculated, and exact $O(\sqrt{n})$ iteration complexity, but not as efficient as the heuristic Mehrotra predictor-corrector IPM algorithm (which sacrifices global convergence). This letter proposes a new predictor-corrector IPM algorithm that preserves the same certified $O(\sqrt{n})$ iteration complexity while achieving a $5\times$ speedup over [1]. Numerical experiments and codes that validate these results are provided.
title A Time-certified Predictor-corrector IPM Algorithm for Box-QP
topic Optimization and Control
url https://arxiv.org/abs/2510.04467