C Code Generation Applied to Nonlinear Model Predictive Control for an Artificial Pancreas
D. Boiroux, J.B. Jørgensen
Technical University of Denmark
Abstract
This paper presents a method to generate C code from MATLAB code applied to a nonlinear model predictive control (NMPC) algorithm. The C code generation uses the MATLAB Coder Toolbox. It can drastically reduce the time required for development compared to a manual porting of code from MATLAB to C, while ensuring a reliable and fairly optimized code. We present an application of code generation to the numerical solution of nonlinear optimal control problems (OCP). The OCP uses a sequential quadratic programming algorithm with multiple shooting and sensitivity computation. We consider the problem of glucose regulation for people with type 1 diabetes as a case study. The average computation time when using generated C code is 0.21 s (MATLAB: 1.5 s), and the maximum computation time when using generated C code is 0.97 s (MATLAB: 5.7 s). Compared to the MATLAB implementation, generated C code can run in average more than 7 times faster.
Full paper
Session
Model Predictive Control (Lecture)
Reference
Boiroux, D.; Jørgensen, J.B.: C Code Generation Applied to Nonlinear Model Predictive Control for an Artificial Pancreas. Editors: Fikar, M. and Kvasnica, M., In Proceedings of the 2017 21st International Conference on Process Control (PC), Štrbské Pleso, Slovakia, June 6 – 9, 327–332, 2017.
BibTeX
| @inProceedings{pc2017-102, | ||
| author | = { | Boiroux, D. and Jørgensen, J.B.}, |
| title | = { | C Code Generation Applied to Nonlinear Model Predictive Control for an Artificial Pancreas}, |
| booktitle | = { | Proceedings of the 2017 21st International Conference on Process Control (PC)}, |
| year | = { | 2017}, |
| pages | = { | 327-332}, |
| editor | = { | Fikar, M. and Kvasnica, M.}, |
| address | = { | \v{S}trbsk\'e Pleso, Slovakia}} |