A Dual Active-Set Solver for Embedded Quadratic Programming Using Recursive LDL' Updates

TL;DR

This paper introduces a dual active-set quadratic programming solver optimized for embedded model predictive control, emphasizing efficiency, simplicity, and robustness through recursive LDL' updates and proximal-point iterations.

Contribution

It presents a novel dual active-set solver with recursive LDL' updates, designed for embedded applications, featuring predictable complexity and robustness to ill-conditioning.

Findings

Efficient solver suitable for embedded MPC

Easily warm-started and simple to implement

Handles ill-conditioned problems robustly

Abstract

In this paper we present a dual active-set solver for quadratic programming which has properties suitable for use in embedded model predictive control applications. In particular, the solver is efficient, can easily be warm-started, and is simple to code. Moreover, the exact worst-case computational complexity of the solver can be determined offline and, by using outer proximal-point iterations, ill-conditioned problems can be handled in a robust manner.