A comparison of interior point and active set methods for FPGA implementation of model predictive control

Mark S K LAU, S. P. Yue, K. V. Ling, J. M. Maciejowski

Research output: Chapter in book/report/conference proceedingConference contributionpeer-review

41 Citations (Scopus)

Abstract

A key component of model predictive control (MPC) is the solving of quadratic programming (QP) problems. Interior point method (IPM) and active set method (ASM) are the most commonly employed approaches for solving general QP problems. This paper compares several performance aspects of the two methods when they are implemented on a FPGA for MPC applications. We compare the computational complexity, storage, convergence speed, and some practical implementation issues. We find that, in general, ASM gives lower complexity and converges faster when the numbers of decision variables and constraints are small. Otherwise, IPM should be a better choice due to its scalability. We also note occasional instability of both IPM and ASM when they are implemented in our FPGA, which uses single precision floating point arithmetic. The instability is mainly due to numerical error, which is found to be more serious in ASM than IPM in our current implementations.

Original languageEnglish
Title of host publication2009 European Control Conference, ECC 2009
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages156-161
Number of pages6
ISBN (Electronic)9783952417393
DOIs
Publication statusPublished - 26 Mar 2014
Event2009 10th European Control Conference, ECC 2009 - Budapest, Hungary
Duration: 23 Aug 200926 Aug 2009

Publication series

Name2009 European Control Conference, ECC 2009

Conference

Conference2009 10th European Control Conference, ECC 2009
Country/TerritoryHungary
CityBudapest
Period23/08/0926/08/09

Scopus Subject Areas

  • Control and Systems Engineering

Fingerprint

Dive into the research topics of 'A comparison of interior point and active set methods for FPGA implementation of model predictive control'. Together they form a unique fingerprint.

Cite this