Release Notes for Model

Predictive Control Toolbox

How to Contact MathWorks
www.mathworks.com Web
comp.soft-sys.matlab Newsgroup
www.mathworks.com/contact_TS.html Technical Support
[email protected] Product enhancement suggestions
[email protected] Bug reports
[email protected] Documentation error reports
[email protected] Order status, license renewals, passcodes
[email protected] Sales, pricing, and general information
508-647-7000 (Phone)
508-647-7001 (Fax)
The MathWorks, Inc.
3 Apple Hill Drive
Natick, MA 01760-2098
For contact information about worldwide offices, see the MathWorks Web site.
Release Notes for Model Predictive Control Toolbox
COPYRIGHT 20052013 by The MathWorks, Inc.
The software described in this document is furnished under a license agreement. The software may be used
or copied only under the terms of the license agreement. No part of this manual may be photocopied or
reproduced in any form without prior written consent from The MathWorks, Inc.
FEDERAL ACQUISITION: This provision applies to all acquisitions of the Program and Documentation
by, for, or through the federal government of the United States. By accepting delivery of the Program
or Documentation, the government hereby agrees that this software or documentation qualifies as
commercial computer software or commercial computer software documentation as such terms are used
or defined in FAR 12.212, DFARS Part 227.72, and DFARS 252.227-7014. Accordingly, the terms and
conditions of this Agreement and only those rights specified in this Agreement, shall pertain to and govern
the use, modification, reproduction, release, performance, display, and disclosure of the Program and
Documentation by the federal government (or other entity acquiring for or through the federal government)
and shall supersede any conflicting contractual terms or conditions. If this License fails to meet the
governments needs or is inconsistent in any respect with federal procurement law, the government agrees
to return the Program and Documentation, unused, to The MathWorks, Inc.
Trademarks
MATLAB and Simulink are registered trademarks of The MathWorks, Inc. See
www.mathworks.com/trademarks for a list of additional trademarks. Other product or brand
names may be trademarks or registered trademarks of their respective holders.
Patents
MathWorks products are protected by one or more U.S. patents. Please see
www.mathworks.com/patents for more information.
Contents
R2013a
R2012b
R2012a
Run-Time Preview of Reference and Measured Disturbance
Signals with MPC Controller Block . . . . . . . . . . . . . . . . . 6
R2011b
C Code Generation Improvements for All Targets with MPC
Controller Block . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Faster QP Solver Algorithm for Improving MPC Controller
Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Run-Time Weight Tuning and Constraint Softening for
MPC Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Run-Time Monitoring of MPC Controller Performance
to Detect When an Optimal Solution Cannot Not Be
Found . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
review Command for Diagnosing Issues with MPC
Controller Parameters That Could Lead to Run-Time
Failures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
mpcmove Returns Aligned Time Horizons for Optimal
Control, Predicted Output and Estimated State . . . . . . . 15
Functionality Being Removed or Changed . . . . . . . . . . . . . 16
iii
R2011a
Support for Custom Constraints on MPC Controller Inputs
and Outputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Ability to Specify Terminal Constraints and Weights on
MPC Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Ability to Access Optimal Cost and Optimal Control
Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
R2010b
No New Features or Changes
R2010a
New Ability to Analyze SISO Generalized Predictive
Controllers (GPC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
R2009b
R2009a
New Sensitivity Analysis to Determine Effect of Weights on
Tuning MPC Controllers . . . . . . . . . . . . . . . . . . . . . . . . . 28
R2008b
New Multiple MPC Controllers Block in the Model
Predictive Control Toolbox Simulink Library . . . . . . . . . 30
iv Contents
Tested Code Generation Support for Real-Time Workshop
Target Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Ability to Design Controllers with Time-Varying Weights
and Constraints Using the GUI . . . . . . . . . . . . . . . . . . . . 32
R2008a
No New Features or Changes
R2007b
New Option for Specifying Time-Varying Constraints . . . . 36
Ability to Specify Nondiagonal Q and R Weight Matrices in
the Cost Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
R2007a
R2006b
No New Features or Changes
R2006a
Bumpless Transfer Added to MPC Block . . . . . . . . . . . . . . . 44
New Bumpless Transfer Demo . . . . . . . . . . . . . . . . . . . . . . . 45
v
R14SP3
No New Features or Changes
R14SP2
No New Features or Changes
vi Contents
R2013a
Version: 4.1.2
New Features: No
Bug Fixes: Yes
1
R2012b
Version: 4.1.1
New Features: No
Bug Fixes: Yes
3
R2012a
Version: 4.1
New Features: Yes
Bug Fixes: Yes
5
R2012a
Run-Time Preview of Reference and Measured
Disturbance Signals with MPC Controller Block
Compatibility Considerations: Yes
This release introduces the ability to preview signals by using the ref and md
inports of the MPC Controller block and the Multiple MPC Controllers block.
The ref inport now accepts an N-by-Ny signal, where N is the number of
previewing steps and Ny is the number of plant outputs.
The md inport now accepts an N-by-Nmd signal, where N is the number of
previewing steps and Nmd is the number of measured disturbances.
You cannot preview if the input signal is a vector, unless Ny or Nmd, as
appropriate, is 1.
For more information, see:
Improving Control Performance with Look-Ahead (Previewing)
Chemical Reactor with Multiple Operating Points
Compatibility Considerations
In the current release, if you have models with the MPC Controller block or
the Multiple MPC Controllers block, you will see a warning if your blocks
contain:
A custom reference signal specified in the MATLAB

workspace.
A custom disturbance signal specified in the MATLAB workspace.
Custom Reference Signal Specified in MATLAB Workspace
You must clear this warning. If you ignore the warning, the block will assume
that the ref signal is zero. This behavior is equivalent to leaving the ref
inport unconnected.
Without Look-Ahead (Previewing) Option. To eliminate this warning:
1 Add a From Workspace block to your model.
6
Run-Time Preview of Reference and Measured Disturbance Signals with MPC Controller Block
2 Specify your reference signal variable name as the Data parameter of
the From Workspace block.
3 Connect the output of the From Workspace block to the ref inport of the
MPC Controller block or the Multiple MPC Controllers block.
With Look-Ahead (Previewing) Option. To eliminate this warning:
1 Copy the Reference Previewer block from the mpc_preview model
and place it in your model. See Improving Control Performance with
Look-Ahead (Previewing) for more information.
2 Specify your reference signal variable name as the Signal parameter of
the Reference Previewer block. Also specify appropriate values for the
Sampling time and Number of previewing steps parameters.
3 Connect the output of the Reference Previewer block to the ref inport of
the MPC Controller block or the Multiple MPC Controllers block.
Custom Disturbance Signal Specified in MATLAB Workspace
You must clear this warning. If you ignore the warning, the block will assume
that the md signal is zero. This behavior is equivalent to leaving the md inport
unconnected.
Without Look-Ahead (Previewing) Option. To eliminate this warning:
1 Add a From Workspace block to your model.
2 Specify your disturbance signal variable name as the Data parameter of
the From Workspace block.
3 Connect the output of the From Workspace block to the md inport of the
MPC Controller block or the Multiple MPC Controllers block.
With Look-Ahead (Previewing) Option. To eliminate this warning:
1 Copy the Measured Disturbance Previewer block from the mpc_preview
model, and place it in your model. See Improving Control Performance
with Look-Ahead (Previewing) for more information.
2 Specify your measured disturbance signal variable name as the Signal
parameter of the Reference Previewer block. Also specify appropriate
values for the Sampling time and Number of previewing steps
parameters.
7
R2012a
3 Connect the output of the Measured Disturbance Previewer block to the
md inport of the MPC Controller block or the Multiple MPC Controllers
block.
8
R2011b
Version: 4.0
New Features: Yes
Bug Fixes: Yes
9
R2011b
C Code Generation Improvements for All Targets
with MPC Controller Block
The MPC Controller block has been re-implemented using a MATLAB
Function block and now supports code generation for all Simulink

Coder
targets.
For more information, see Code Generation with Simulink Coder.
10
Faster QP Solver Algorithm for Improving MPC Controller Performance
Faster QP Solver Algorithm for Improving MPC
Controller Performance
This release implements a new quadratic problem (QP) solver that uses the
KWIK algorithm. KWIK is faster and more numerically robust than the
previous solver for ill-conditioned QP problems. You can use this solver
without default constraints on decision variables.
For more information, see MPC QP Solver.
11
R2011b
Run-Time Weight Tuning and Constraint Softening
for MPC Controller
This release introduces three new run-time tuning parameters for the MPC
Controller block:
Weights on plant outputs
Weights on manipulated variables rate
Weight on overall constraints softening
You can use these parameters to tune the weights on plant outputs,
manipulated variables rate, and overall constraint softening. These
capabilities are available in real time, without redesigning or re-implementing
the MPC controller, and help adjust the controller performance.
For more information, see Tuning Controller Weights.
You can also use an mpcmoveopt object as an input to mpcmove to tune the
weights and constraints.
For more information, see the following:
Switching Controllers Based on Optimal Costs
Varying Input and Output Constraints
12
Run-Time Monitoring of MPC Controller Performance to Detect When an Optimal Solution Cannot Not Be Found
Run-Time Monitoring of MPC Controller Performance
to Detect When an Optimal Solution Cannot Not Be
Found
This release introduces a new outport parameterOptimization status in
the MPC Controller block. You can use this outport to monitor the status of
the optimization and take the necessary action when an optimal solution
cannot be found. For more information, see Monitoring Optimization Status
to Detect Controller Failures.
You can also use the Info.QPCode field of the output of mpcmove to monitor
the status of the optimization.
For more information, see the mpcmove reference page.
13
R2011b
review Command for Diagnosing Issues with MPC
Controller Parameters That Could Lead to Run-Time
Failures
You can now use review to detect potential stability and robustness issues
(both offline and at run time) with an MPC Controller design. The following
aspects of the system are inspected:
Stability of the model predictive controller and the closed loop
Potential for contradictory settings in the specified constraints and
mitigation of an ill-conditioned QP problem by softening constraints
Validity of QP Hessian matrix
Use this command before implementing the MPC Controller, in conjunction
with simulation.
For more information, see the following:
review reference
Reviewing Model Predictive Controller Design for Potential Stability and
Robustness Issues.
14
mpcmove Returns Aligned Time Horizons for Optimal Control, Predicted Output and Estimated State
mpcmove Returns Aligned Time Horizons for Optimal
Control, Predicted Output and Estimated State
mpcmove now returns Info with a time horizon of t=k,...,k+p, where k is the
current time and p is the prediction horizon for the following fields:
Info.Uopt Optimal manipulated variable adjustments
Info.Yopt Predicted output
Info.Xopt Predicted state
Info.Topt Time horizon
You can now plot Info.Uopt,Info.Yopt and Info.Xopt using Info.Topt as
the time vector.
For more information, see the mpcmove reference page.
15
R2011b
Functionality Being Removed or Changed
Compatibility Considerations: Yes
Functionality What Happens
When You Use This
Functionality?
Use This Instead Compatibility
Considerations
getmpcdata Still runs
get
getconstraint
getestim
getindist
getoutdist
Not applicable
pack Still runs Not applicable Not applicable
qpdantz Warns quadprog (requires
Optimization
Toolbox)
Replace all instances
of qpdantz with
quadprog.
setmpcdata Still runs
set
setconstraint
setestim
setindist
setoutdist
Not applicable
16
R2011a
Version: 3.3
New Features: Yes
Bug Fixes: Yes
17
R2011a
Support for Custom Constraints on MPC Controller
Inputs and Outputs
In addition to upper and lower bounds, you can now specify constraints on
linear combinations of an MPC controller inputs (u(t)) and outputs (y(t)).
Specify custom constraints, such as u1 + u2 < 1 or u + y < 2, in the mpc object
using setconstraint.
For more information, see:
Custom Constraints on Inputs and Outputs
Custom Constraints in a Blending Process
MPC Control with Constraints on a Combination of Input and Output
Signals demo
18
Ability to Specify Terminal Constraints and Weights on MPC Controller
Ability to Specify Terminal Constraints and Weights
on MPC Controller
You can now specify weights and constraints on the terminal predicted states
of an MPC controller.
Using terminal weights, you can achieve infinite horizon control. For
example, you can design an unconstrained MPC controller that behaves in
exactly the same way as a Linear-Quadratic Regulator (LQR). You can use
terminal constraints as an alternative way to achieve closed-loop stability by
defining a terminal region.
You can specify both weights and constraints using the setterminal
command.
For more information, see:
Terminal Weights and Constraints
Using Terminal Penalty to Provide LQR Performance
Implementing Infinite-Horizon LQR by Setting Terminal Weights in a
Finite-Horizon MPC Formulation demo
19
R2011a
Ability to Access Optimal Cost and Optimal Control
Sequence
This release introduces two new parameters Enable optimal cost outport
and Enable control sequence outport in the MPC Controller block. Using
these parameters, you can access the optimal cost and control sequence
along the prediction horizon. This information helps you analyze control
performance.
You can also access the optimal cost and control sequence programmatically
using the new Cost and Yopt fields, respectively, of the structure info
returned by mpcmove.
For more information on using optimal cost and control sequence, see the
following demos:
MPC Control with Input Quantization Based on Comparing the Optimal
Costs
Analysis of Control Sequences Optimized by MPC on a Double Integrator
System
20
R2010b
Version: 3.2.1
New Features: No
Bug Fixes: No
No New Features or Changes
21
R2010a
Version: 3.2
New Features: Yes
Bug Fixes: Yes
23
R2010a
New Ability to Analyze SISO Generalized Predictive
Controllers (GPC)
You can now use gpc2mpc to convert your SISO GPC controller to an MPC
controller. Analyze and simulate the resulting MPC controller using available
Model Predictive Control Toolbox commands.
For more information, see the gpc2mpc reference page.
24
R2009b
Version: 3.1.1
New Features: No
Bug Fixes: Yes
25
R2009a
Version: 3.1
New Features: Yes
Bug Fixes: Yes
27
R2009a
New Sensitivity Analysis to Determine Effect of
Weights on Tuning MPC Controllers
You can now perform sensitivity analysis to determine the effect of weights
on the closed-loop performance of your system. You can perform sensitivity
analysis using the following:
MPC Tuning Advisor. See Tuning Advisor in the Model Predictive Control
Users Guide.
sensitivity command. See the sensitivity reference page.
28
R2008b
Version: 3.0
New Features: Yes
Bug Fixes: Yes
29
R2008b
New Multiple MPC Controllers Block in the Model
Predictive Control Toolbox Simulink Library
You can now use the Multiple MPC Controllers block in Simulink software
to control a nonlinear process over a range of operating points. You include
an MPC controller for each operating point in the Multiple MPC Controllers
block and specify switching between these controllers in real-time based on
the input scheduling signal to the block. If you need to change the design of
a specific controller, you can open the MPC Design Tool GUI directly from
the Multiple MPC Controllers block.
During model simulation, Model Predictive Control Toolbox provides
bumpless transfer when the system transitions between operating points.
To learn more about configuring the new block, see the Multiple MPC
Controllers block reference page.
30
Tested Code Generation Support for Real-Time Workshop

Target Systems
Tested Code Generation Support for Real-Time
Workshop Target Systems
After designing an MPC controller in Simulink software using the MPC
Controller block, you can use Real-Time Workshop

software to build this

controller and deploy it to the following target systems for real-time control:
Generic Real-Time Target
Real-Time Workshop Embedded Coder
Real-Time Windows Target
Rapid Simulation Target
Target Support Package FM5
xPC Target
dSpace Target
Target for Infineon TriCore
The following target systems are either not supported or not recommended
because they result in significant performance issues:
Embedded Target for TI C2000 DSP
Embedded Target for TI C6000 DSP
Target Support Package IC1 (for Infineon C166)
Tornado (VxWorks) Real-Time Target
Note The Multiple MPC Controllers block has not been tested with the
target systems supported by Real-Time Workshop software.
31
R2008b
Ability to Design Controllers with Time-Varying
Weights and Constraints Using the GUI
While you design an MPC controller using the MPC Design Tool graphical
user interface (GUI), you can specify time-varying weights and constraints for
manipulated variables, rate of change of manipulated variables, and output
variables. In the previous version, you could only specify the time-varying
weights and constraints at the command line.
Furthermore, you can load an MPC controller with time-varying information
from the command line into the MPC Design Tool GUI.
To learn more about the new options in the MPC Design Tool GUI, see the
Model Predictive Control Toolbox documentation.
32
R2008a
Version: 2.3.1
New Features: No
Bug Fixes: No
No New Features or Changes
33
R2007b
Version: 2.3
New Features: Yes
Bug Fixes: No
35
R2007b
New Option for Specifying Time-Varying Constraints
You can now configure the Model Predictive Controller block in Simulink to
accept time-varying constraint signals that are generated by other blocks. To
add inports to which you can connect time-varying constraint specifications,
select the new Enable input port for input and output limits check box
in the MPC Controller block. See also the mpcvarbounds demo.
In the previous version, you could only specify the constraints during the
design phase and these contraints remained constant for the duration of the
simulation.
For more information about the new Enable input port for input and
output limits check box in the Model Predictive Controller block, see the
MPC Controller block reference page.
36
Ability to Specify Nondiagonal Q and R Weight Matrices in the Cost Function
Ability to Specify Nondiagonal Q and R Weight
Matrices in the Cost Function
You can now specify off-diagonal weights in the cost function. In the previous
release, only diagonal Q and R matrices were supported.
To learn more about specifying off-diagonal weights, see the discussion about
weights in the MPC Controller block reference pages.
To access a new demo that shows how to use nondiagonal weight matrices,
type the following command at the MATLAB prompt:
showdemo('mpcweightsdemo')
37
R2007a
Version: 2.2.4
New Features: No
Bug Fixes: Yes
39
R2006b
Version: 2.2.3
New Features: No
Bug Fixes: No
No New Features or Changes
41
R2006a
Version: 2.2.2
New Features: Yes
Bug Fixes: No
43
R2006a
Bumpless Transfer Added to MPC Block
Bumpless transfer between manual and automatic operation or from one
controller to another has been added to the Model Predictive Controller block
in Simulink. This block now allows feedback of the true manipulated variable
signals, which allows the controller to maintain an accurate state estimate
during periods when its calculated adjustments are not being sent to the
plant. For example, the controllers output might be ignored during a startup
period or during temporary intervention by a (simulated) plant operator.
If the controller assumes that its adjustments are being implemented (the
default behavior), its state estimate will be incorrect, leading to a bump
when the controller is reconnected to the plant. A tutorial example has been
added to the documentation.
44
New Bumpless Transfer Demo
New Bumpless Transfer Demo
A new demo illustrating bumpless transfer has been added to the toolbox.
45
R14SP3
Version: 2.2.1
New Features: No
Bug Fixes: No
No New Features or Changes
47
R14SP2
Version: 2.2
New Features: No
Bug Fixes: No
No New Features or Changes
49