setCustomSolver

Configures an MPC object to use the QP solver from Optimization Toolbox as a custom solver

Syntax

setCustomSolver(mpcobj,'quadprog')

setCustomSolver(mpcobj,'none')

Description

setCustomSolver(mpcobj,'quadprog') configures mpcobj to use quadprog from Optimization Toolbox™ as a custom QP solver for both simulation and code generation. Specifically, this syntax generates, in the current folder, the files mpcCustomSolver.m and mpcCustomSolverCodeGen.m, which internally call the active-set quadprog (Optimization Toolbox) solver. It then sets mpcobj.Optimizer.CustomSolver and mpcobj.Optimizer.CustomSolverCodeGen to true.

setCustomSolver(mpcobj,'none') sets mpcobj.Optimizer.CustomSolver and mpcobj.Optimizer.CustomSolverCodeGen to false, thereby reverting mpcobj back to use the built-in algorithm specified in mpcobj.Optimizer.Algorithm for both simulation and code generation.

Examples

Set quadprog as Custom MPC Solver Using setCustomSolver

This example shows how to use the setCustomSolver function to automatically configure an mpc object to use the Optimization Toolbox™ quadprog function as custom MPC solver for both simulation and code generation.

Create an mpc object.

mpcobj = mpc(tf(1,[2 1],0.1));

-->The "PredictionHorizon" property is empty. Assuming default 10.
-->The "ControlHorizon" property is empty. Assuming default 2.
-->The "Weights.ManipulatedVariables" property is empty. Assuming default 0.00000.
-->The "Weights.ManipulatedVariablesRate" property is empty. Assuming default 0.10000.
-->The "Weights.OutputVariables" property is empty. Assuming default 1.00000.

As a default, the controller is set to use the active-set solver for both simulation and code generation.

mpcobj.Optimizer

ans = struct with fields:
                Algorithm: 'active-set'
         ActiveSetOptions: [1x1 struct]
     InteriorPointOptions: [1x1 struct]
      MixedIntegerOptions: [1x1 struct]
             MinOutputECR: 0
    UseSuboptimalSolution: 0
             CustomSolver: 0
      CustomSolverCodeGen: 0

Configure mpcobj to Use quadprog as Custom Solver

To set the quadprog function as custom MPC solver for both simulation and code generation, call setCustomSolver with 'quadprog' as second argument.

setCustomSolver(mpcobj,'quadprog')

The function generates, in the current folder, the files mpcCustomSolver.m and mpcCustomSolverCodeGen.m. To display the MATLAB files in the current folder, use the ls command.

ls *.m

mpcCustomSolverCodeGen.m  mpcCustomSolver.m

To display the content of mpcCustomSolver.m, and mpcCustomSolverCodeGen.m use the type command. Both files internally call quadprog, which is configured to use the active-set solver, as other algorithms are not supported.

type mpcCustomSolver

function [x, status] = mpcCustomSolver(H, f, A, b, x0)
% "mpcCustomSolver" enables using "quadprog" from Optmization Toolbox 
% as a custom QP solver with linear MPC controller for simulation.

%% Specify solver algorithm and options
options = optimoptions('quadprog','Algorithm','active-set');
if coder.target('MATLAB')
    options.Display = 'none';  
end
%% Process solver inputs
% Use -A and -b in "quadprog" because MPC QP uses Ax>=b instead
A_custom = -A;
b_custom = -b;
% ensure Hessian is symmetric
H = (H+H')/2; 
%% Call "quadprog"
[x, ~, exitflag, output] = quadprog(H, f, A_custom, b_custom, [], [], [], [], x0, options);
%% Converts exit flag to MPC "status"
switch exitflag
    case 1
        status = output.iterations;
    case 0
        status = 0;
    case -2
        status = -1;
    otherwise
        status = -2;
end
%% If "quadprog" fails to find a solution, set x to the initial guess
if status <= 0
    x = x0;
end

type mpcCustomSolverCodeGen.m

function [x, status] = mpcCustomSolverCodeGen(H, f, A, b, x0)
% "mpcCustomSolverCodeGen" enables using "quadprog" from Optmization 
% Toolbox as a custom QP solver with linear MPC controller for code generation.

%#codegen
%% Specify solver algorithm (must be "active-set") and options
options = optimoptions('quadprog','Algorithm','active-set');
if coder.target('MATLAB')
    options.Display = 'none';  
end
%% Process solver inputs
% Use -A and -b in "quadprog" because MPC QP uses Ax>=b instead
A_custom = -A;
b_custom = -b;
% ensure Hessian is symmetric
H = (H+H')/2; 
%% Call "quadprog"
[x, ~, exitflag, output] = quadprog(H, f, A_custom, b_custom, [], [], [], [], x0, options);
%% Converts exit flag to MPC "status"
switch exitflag
    case 1
        status = output.iterations;
    case 0
        status = 0;
    case -2
        status = -1;
    otherwise
        status = -2;
end
%% If "quadprog" fails to find a solution, set x to the initial guess
if status <= 0
    x = x0;
end

The setCustomSolver function also sets mpcobj.Optimizer.CustomSolver and mpcobj.Optimizer.CustomSolverCodeGen to true, thereby setting up the mpcobj object to use the custom solver in the related files for simulation and code generation.

mpcobj.Optimizer

ans = struct with fields:
                Algorithm: 'active-set'
         ActiveSetOptions: [1x1 struct]
     InteriorPointOptions: [1x1 struct]
      MixedIntegerOptions: [1x1 struct]
             MinOutputECR: 0
    UseSuboptimalSolution: 0
             CustomSolver: 1
      CustomSolverCodeGen: 1

Revert mpcobj to Use a Built-In Solver

To revert mpcobj back to use a built in solver, call setCustomSolver function with 'none' as second argument.

setCustomSolver(mpcobj,'none')

This sets mpcobj.Optimizer.CustomSolver and mpcobj.Optimizer.CustomSolverCodeGen to false.

mpcobj.Optimizer

ans = struct with fields:
                Algorithm: 'active-set'
         ActiveSetOptions: [1x1 struct]
     InteriorPointOptions: [1x1 struct]
      MixedIntegerOptions: [1x1 struct]
             MinOutputECR: 0
    UseSuboptimalSolution: 0
             CustomSolver: 0
      CustomSolverCodeGen: 0

The controller will now use the active-set built in solver for both simulation and code generation.

Input Arguments

mpcobj — MPC controller

MPC controller object

MPC controller, specified as an MPC controller object. Use the mpc command to create the MPC controller.

Version History

Introduced in R2021b