BnB-DAQP: A Mixed-Integer QP Solver for Embedded Applications
Optimization and Control
2022-11-24 v1
Abstract
We propose a mixed-integer quadratic programming (QP) solver that is suitable for use in embedded applications, for example, hybrid model predictive control (MPC). The solver is based on the branch-and-bound method, and uses a recently proposed dual active-set solver for solving the resulting QP relaxations. Moreover, we tailor the search of the branch-and-bound tree to be suitable for embedded applications on limited hardware; we show, for example, how a node in the branch-and-bound tree can be represented by only two integers. The embeddability of the solver is shown by successfully running MPC of an inverted pendulum on a cart with contact forces on an MCU with limited memory and computing power.
Cite
@article{arxiv.2211.12902,
title = {BnB-DAQP: A Mixed-Integer QP Solver for Embedded Applications},
author = {Daniel Arnström and Daniel Axehill},
journal= {arXiv preprint arXiv:2211.12902},
year = {2022}
}