ntnu

MPC and Optimization, Tor Arne Johansen, NTNU

Fast Model Predictive Control

Traditionally Model Predictive Control (MPC) is implemented via real-time optimization. Our objective is to find explicit state feedback laws that are equivalent to MPC and thereby allow MPC to be implemented simply via the evaluation of a piecewise linear function. Our results show that this is possible for applications of moderate size. The benefits of this are

  • MPC can be implemented at high sampling rates (milliseconds or less) on inexpensive computers using fixed-point arithmetic
  • MPC software reliability can be easily verified and applied in safety-critical applications

Multi-parametric programming:

Online computations in explicit MPC

Applications of explicit MPC

  • L. Feng, C. R. Gutvik, T. A. Johansen, D. Sui, A. O. Brubakk, Approximate Explicit Nonlinear Receding Horizon Control for Decompression of Divers, IEEE Trans. Control Systems Technology, accepted, 2011
  • C. R. Gutvik, T. A. Johansen, A. O. Brubakk, Optimal Decompression of Divers - Procedures for constraining predicted bubble growth, IEEE Control Systems Magazine, Vol. 31, No. 1, pp. 19-28, 2011; DOI:10.1109/MCS.2010.939141 
  • L. Feng, C. R. Gutvik, T. A. Johansen, Optimal Decompression Through Multi-parametric Nonlinear Programming, IFAC NOLCOS, Bologna, Italy, 2010
  • A.Grancharova, T. A. Johansen, Design and Comparison of Explicit Model Predictive Controllers for an Electropneumatic Clutch Actuator Using On/Off Valves, IEEE/ASME Trans. Mechatronics, accepted, 2011
  • Grancharova, T. A. Johansen and J. Kocijan, Explicit model predictive control of gas-liquid separation plant, Computers and Chemical Engineering, Vol. 28, 2004

Other MPC formulation

Optimizing constrained control allocation

Control allocation deals with the task of coordinating a redundant set of actuators to produce a totalt effort, taking into account actuator and operational constraints and objective functions.

Key publications