US 2015/0145558 A1 · METHOD FOR CREATING DIGITAL CIRCUITS OF A FEEDBACK CONTROL SYSTEM THAT IMPLEMENTS AN APPROXIMATION TECHNIQUE FOR MODEL PREDICTIVE CONTROL (MPC)

Abstract

Method for creating digital circuits of an MPC controller (2), which implements an approximation technique for Model Predictive Control (MPC), in which a quadratic programming (QP) or linear programming (LP) optimization problem is formulated by starting from a model defined over a set of states of a state (x), wherein the said set of states is partitioned into simplices identified by vertices (v), and wherein said method comprises the steps of a) compute a solution (w*) of the optimization problem and define a control law (u(x)); b) check the stability under the control law (u(x)); c) synthesize a digital circuit by starting from the control law (u(x)) and the vertices (vi).

Claims15

Dependent claims

Method for creating digital circuits comprised in an MPC controller, which implements an approximation technique for Model Predictive Control (MPC), in which a quadratic programming or linear programming optimization problem is formulated based on a linear state-space dynamical model of a physical process having a state x, and provides a control law (u(x)) defined over a partition that contains simplices having vertices (vi), where i ranges from 1 to Nv, and wherein the vertices (vi) are comprised in a given set of states of the physical process, wherein said method comprises the following steps:

a. compute a solution (w*) of the optimization problem and define the control law (u(x));
b. check the stability of a closed-loop system under the control law u(x), wherein the closed-loop system comprises the MPC controller and the physical process;
and
c. synthesize a digital circuit by starting from the control law (u(x)) and the vertices (vi).

1 more element

Circuit architecture for computing a value of at least one piecewise-affine simplicial (PWAS) function (fPWL(z)), comprising an address generator, a memory, and a pt-generator, wherein said address generator takes coordinates of a point (z) as input, and the point (z) having n dimensions is obtained by resealing a state (x) of a model, and wherein the circuit is able to perform the following steps:

a. locate an hypercube containing the point (z) by using integer parts of the coordinates of the point (z) as input of the address generator;
b. locate which of n! simplices contained in the selected hypercube contains the point (z) by using decimal parts of the coordinates of the point (z) as input of the μ-generator and of the address generator;
and
c. compute the value of the PWAS function (fPWL(z)) in point (z) by linear interpolation of the value of said function (fPWL(z)) in the vertices of the selected simplex;
wherein, both the address generator and the memory are synthesized starting from a control law (u(x)) and vertices (vi), which are computed by a method according to claim 19.

2 more elements

Dependencies are inferred from the claim text.

Method for creating digital circuits of a feedback control system that implements an approximation technique for model predictive control (mpc)

Life of this patent

Method for creating digital circuits of a feedback control system that implements an approximation technique for model predictive control (mpc)

Life of this patent

Abstract

Claims15