The First American Model Predictive Control Summer School

Robust Model Predictive Control

(Part I)

Saša V. Raković, Ph.D. DIC

Two Pending (Secret) Affiliations

University of Wisconsin-Madison,
Madison, Wisconsin, July 27, 2017
Outline

Opening

Model Predictive Control

Robust Model Predictive Control

Dissecting Uncertainty Effects

Tube Model Predictive Control

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 1
Robust Model Predictive Control (Part I)

Opening

Model Predictive Control

Robust Model Predictive Control

Dissecting Uncertainty Effects

Tube Model Predictive Control

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 2
Objectives

MPC under uncertainty

Repetitive decision making process point of view.
Technical aspects.
Computational issues.
Understanding
Uncertainty and its effects on dynamics, constraints and cost.
MPC under uncertainty.

Robert J. Aumann (1930 – )

The Nobel Memorial Prize Laureate in Economics (2005).

Repeated games,
Fundamental work in game theory and mathematics, and
Interesting illustration of Understanding.
Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 3
Robust Model Predictive Control (Part I)

Opening

Model Predictive Control

Robust Model Predictive Control

Dissecting Uncertainty Effects

Tube Model Predictive Control

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 4
Model Predictive Control Paradigm

Goals:
Constraint satisfaction,
Stability, and
Optimized performance.

Tool:
Model predictive control.

Model predictive control (MPC):

Repetitive decision making process (DMP).
Basic DMP is finite horizon optimal control (FHOC).

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 5
Basic DMP = Finite Horizon Optimal Control

Given an integer N ∈ N and a state x ∈ X select predicted sequences of

control actions uN−1 := {u0 , u1 , . . . , uN−1 }, and
controlled states xN := {x0 , x1 , . . . , xN−1 , xN },
which, for each k ∈ {0, 1, . . . , N − 1}, satisfy

xk+1 = f (xk , uk ) with x0 = x,

xk ∈ X,
uk ∈ U, and
xN ∈ Xf ,

PN−1
and which minimize VN (xN , uN−1 ) := k=0 `(xk , uk )+ Vf (xN ) .
(Hereafter, Xf and Vf (·) are terminal constraint set and cost function.)

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 6
Prototype Algorithm and Principal Components
Prototype Algorithm:
At any state x and time t in the control process.
Solve basic DMP to evaluate control u00 (x) and value function VN0 (x).
Implement control action u00 (x).

Principal Components:
Control law u00 (·).
(Possibly set–valued) Feedback implicitly evaluated at current state.
Closed–loop controlled dynamics x + = f (x, u00 (x)).
(Possibly set–valued) Implicitly evaluated at encountered states.
Value function VN0 (·).
Lyapunov certificate for closed–loop controlled dynamics.
Controllability set, the domain of the value function, XN .
Positively invariant set for closed–loop controlled dynamics.
Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 7
Key Facts

Main properties:
MPC law u00 (·) is feedback implicitly evaluated at current state.
Predictions and optimized predictions are, however, open–loop.
Consistently improving and stabilizing (under mild assumptions).
Theoretical implementation:
Mathematical (nonlinear) programming in general case.
Strictly convex programming in most frequent cases.
Practical implementation:
Online optimization.
Offline parameteric optimization and online look–up tables.
Combinations of the online and offline parameteric optimization.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 8
Main Properties
Under mild assumptions on problem setting and terminal conditions
(e.g., regular continuous–compact-ls–continuous setting) design process is:
Well–posed.
(u00 (·), f (·, u00 (·)), VN0 (·) and XN well defined.)
Consistently improving.
(Increasing N decreases VN0 (·) and enlarges XN , etc.)
Positive invariance–inducing.
(Initial feasibility ⇒ ad–infinitum feasibility.)
Stabilizing.
(Ensures stabilization of the desired equilibrium.)
Optimizing.
(Enhances performance of the induced control process.)
Computationally plausible/effective.
(Systematic implementation via standard optimization techniques.)
Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 9
A Few Points

MPC is the best thing since sliced bread!

High impact on diverse parts of academic community.

Versatile applicability and beneficial utility over range of industries.

MPC’s evolution has followed an interesting path.

Many misconceptions, and considerable time to overcome these.

Relatively mature, but still offering considerable challenges.

MPC’s feedback nature and robustness.

A form of feedback due to repetition of open–loop optimal control.

Open–loop optimal control might be fragile (non–robust), and it is not

the most adequate tool for control synthesis under uncertainty.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 10
Uncertainty Sources and Types

Uncertainty Sources:
Modeling. (Approximation, imprecision, etc.)
Sensing. (Imperfection, estimation, etc.)
Computing. (Cyber realm, software limitations, etc.)

Implementing. (Physical realm, hardware limitations, etc.)

...

Uncertainty Types:
Set–membership.
Probabilistic.
Combined.

Systematic analysis of uncertainty and its forms is a topic in its own right!

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 11
Robust Model Predictive Control (Part I)

Opening

Model Predictive Control

Robust Model Predictive Control

Dissecting Uncertainty Effects

Tube Model Predictive Control

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 12
Robust Model Predictive Control Paradigm

Goals:
Robust constraint satisfaction,
Robust stability,
Optimized robust performance, and
Computational practicability.

Tool:
Robust model predictive control.

Robust model predictive control (RMPC):

Repetitive decision making process (DMP).
Basic DMP is finite horizon robust optimal control.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 13
Pivotal Concerns
Convoluted interaction of uncertainty with:

System evolution,
Constraints, and
Performance.

Fragility (non–robustness) of conventional MPC,

Intractability of exact, closed–loop robust MPC,

Insensitivity of naive, open–loop robust MPC,

Conflicting demands:

Strong structural properties and

Online computational tractability.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 14
Value of Information
Scalar system: x + = x + u + w,
Constraints: X = [−1, 1], U = [−3, 3] and W = [−2, 2],
Target Set: Xf = {0}

Objective I Minimax stabilization of Xf .

(Minimax control functions u (·) : X → U.)

Objective II Maximin stabilization of Xf .

(Maximin control functions u (·, ·) : X × W → U.)

Observations: Minimax stabilization not possible, as Xf ⊆ X ⊆ W.

Maximin stabilization possible, take u(x, w) = −x − w.

Explanation: Inequality supa∈A inf b∈B c(a, b) ≤ inf b∈B supa∈A c(a, b).

Our focus is on minimax informational setting.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 15
Predicting Without Uncertainty (x + = x + u)
States xk depend on:
initial state x0 , and
controls u0 , u1 , . . . , uk−1 .

Controls uk depend on:

initial state x0 .

Open–loop optimal control

=
Closed–loop optimal control.

Can use open–loop control sequences!

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 16
Predicting Without Uncertainty (x + = x + u)

In charge of selecting
states {x0 , x1 , . . . , xN }, and
controls {u0 , u1 , . . . , uN−1 }.

These are open–loop, and:

dynamically consistent,
constraint admissible, and
cost optimal.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 16
Predicting Under Uncertainty (x + = x + u + w)
States xk depend on:
initial state x0 ,
controls u0 , u1 , . . . , uk−1 , and
disturbances w0 , w1 . . . , wk−1 .

Controls uk depend on:

current state xk .

Disturbances wk are independent.

Open–loop optimal control

 (often ≺)
Closed–loop optimal control.

Should use closed–loop control policies!

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 17
Predicting Under Uncertainty (x + = x + u + w)
Not in charge of selecting
states {x0 , x1 , . . . , xN }, and
controls {u0 , u1 , . . . , uN−1 }.

These should be closed–loop, and

worst–case:
dynamically consistent,
constraint admissible, and
cost optimal.

In charge of selecting sets of

possible
states {X0 , X1 , . . . , XN }, and
controls {U0 , U1 , . . . , UN−1 }.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 17
Robustness of Conventional MPC

Necessary to consider robustness

properties w.r.t. both cost and
constraints.

MPC is robust in some sense for some

classes of problems. But, MPC is
fragile because of discontinuity of the
value function and its optimizer.

The absence of nominal robustness is

“predicted” by the celebrated Arstein’s
Theorem, and “illustrated” by the
famous Arstein’s circles.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 18
Fragility of Conventional MPC

Key issues (mostly due to state, including terminal, constraints):

Asymptotic stability needs not be a robust property (Teel),

Optimal control of a continuous control system might induce a
discontinuous controlled dynamics, and
Optimal control might be a fragile process itself (Raković).

Message:

There’s no thing such as a free lunch.

Ensure robustness by design rather than hoping to get it for free.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 19
Computational Complexity of Exact Robust MPC

Exact robust MPC employs exact robust optimal control.

Exact robust optimal control is an infinite–dimensional

minimaximization problem taking generic form:

minΠN−1 ∈ΠN−1 (x) maxwN−1 ∈WN VN (x, ΠN−1 , wN−1 ).

Solvable via dynamic programming.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 20
Dynamic Programming Based Robust MPC
Richard E. Bellman (1920 – 1984)

Dynamic Programming

Curse of Dimensionality

Minimax DP Recursion (with boundary conditions V0 (·) := Vf (·) and X0 := Xf )

Max value functions Jk (·):

∀(x, u) ∈ Xk × U, Jk (x, u) = maxw {`(x, u, w) + Vk−1 (f (x, u, w)) : w ∈ W}.
Minimax value functions Vk (·):
∀x ∈ Xk , Vk (x) = minu {Jk (x, u) : u ∈ U ∧ ∀w ∈ W, f (x, u, w) ∈ Xk−1 }.

Minimax optimal control laws uk (·) are the optimizers of the minimax value functions:
∀x ∈ Xk , uk (x) = arg minu {Jk (x, u) : u ∈ U ∧ ∀w ∈ W, f (x, u, w) ∈ Xk−1 }.
Domains of the minimax value functions Xk are the minimax controllability sets:
Xk = F −1 (Xk−1 ) where F −1 (X ) = {x ∈ X : ∃u ∈ U, ∀w ∈ W, f (x, u, w) ∈ X }.
For DP based RMPC, one makes (repetitive) use of (a selection of) uN (·) and VN (·) over XN .

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 21
Dynamic Programming Based Robust MPC
Richard E. Bellman (1920 – 1984)

Dynamic Programming

Curse of Dimensionality

Minimax DP Recursion (with boundary conditions V0 (·) := Vf (·) and X0 := Xf )

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 21
Dynamic Programming Based Robust MPC
Richard E. Bellman (1920 – 1984)

Dynamic Programming

Curse of Dimensionality

Minimax DP Recursion (with boundary conditions V0 (·) := Vf (·) and X0 := Xf )

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 21
Closed–Loop Robust MPC

System:

x+ = x + u + w

Uncertainty:

w ∈ [−1, 1]

Prediction horizon:

N=4

Closed–Loop (or brute force scenarios based) RMPC is clearly intractable!

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 22
Open–Loop Robust MPC

System:

x+ = x + u + w

Uncertainty:

w ∈ [−1, 1]

Prediction horizon:

N=4

Open–Loop (or a careless man crossing street) RMPC is clearly insensitive!

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 23
Breaking Down Complexity and Robust MPC Consensus
What robust MPC a rational and intelligent man should be happy with?
Improved computability w.r.t. DP based and closed–loop RPMC.
Improved sensibility w.r.t. conventional and open–loop RMPC.
And anything else on top of that as a bonus.

Need for generalized:

Control and state predictions,
Constraint satisfaction,
Stability notions, and
Performance criteria,

with the aim to:

retain structural properties and
ensure computational practicability.
Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 24
Robust Model Predictive Control (Part I)

Opening

Model Predictive Control

Robust Model Predictive Control

Dissecting Uncertainty Effects

Tube Model Predictive Control

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 25
Local Dynamics Under Uncertainty

System: x + = f (x, u, w),

Constraints: x ∈ X, u ∈ U and w ∈ W,
Control function: πf (·).

State set–dynamics X + = F(X , πf ),

Control set–dynamics U = G(X , πf ).

State sets Xk ,
Control sets Uk .

Question I What is an equilibrium under uncertainty?

Question II What is a domain of attraction under uncertainty?

State map F(X , πf ) := {f (x, u, w) : x ∈ X , u ∈ πf (x), w ∈ W},

Control map G(X , πf ) := {u : x ∈ X , u ∈ πf (x)}.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 26
Local Behaviour Under Uncertainty

Positive Invariance and Attractivity

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 27
Equilibrium Under Uncertainty
Generalized dynamics are set–dynamics.
Generalized state is set of states.
Generalized control is set of controls.

Generalized equilibrium is a set.

(Corresponds to the minimal RPI set.)
Generalized equilibrium pair is a pair of sets.

Without Uncertainty Under Uncertainty

Fixed point equation: Fixed point set equation:
x = f (x, κf (x)) (e.g., x = 0). X = F(X , πf ).
Admissibility: Admissibility:
x ∈ X ∧ u := κf (x) ∈ U. X ⊆ X ∧ U := G(X , πf ) ⊆ U.
Generalized equilibrium is unique in compact–contractive setting.
Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 28
Local Domain of Attraction Under Uncertainty
Generalized dynamics are set–dynamics.
Generalized state is set of states.
Generalized control is set of controls.

Local domain of attraction is a set of sets.

(The union of which is of main interest and
corresponds to the maximal RPI set.)

Without Uncertainty Under Uncertainty

Maximal PI set (in strictly Maximal PI set of sets (in strictly
stable case): Xf . stable case): Xf .
Properties: maximal set s.t. Properties: maximal set of sets s.t.
Xf ⊆ X, G(Xf , κf ) ⊆ U ∧ ∀X ∈ Xf , X ⊆ X, G(X , πf ) ⊆
F(Xf , πf ) ⊆ Xf . U ∧ F(X , πf ) ∈ Xf .

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 29
Global Dynamics Under Uncertainty
With use of control policy ΠN−1 (·) := {π0 (·, ·) , π1 (·, ·) , . . . , πN−1 (·, ·)}.

State set–dynamics Xk+1 = F(Xk , πk ),

Control set–dynamics Uk = G(Xk , πk ).

State sets Xk ,
Control sets Uk .

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 30
Constraints Under Uncertainty
With use of control policy ΠN−1 (·) := {π0 (·, ·) , π1 (·, ·) , . . . , πN−1 (·, ·)}.

Initial state constraints x ∈ X0 ,

State constraints Xk ⊆ X,
Control constraints Uk ⊆ U,

Terminal constraints XN ∈ Xf typically expressed as XN ⊆ Xf .

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 31
Cost Under Uncertainty

Without Uncertainty Under Uncertainty

Stage cost: Stage cost:
`(x, u) := `(x, u) :=
dist(Q, x, {x}) + dist(R, u, {u}). dist(Q, x, X ) + dist(R, u, U).
Terminal cost: Terminal cost:
Vf (x) := dist(P, x, {x}). Vf (x) := dist(P, x, X ).

Distance function
dist(S, p, S) := inf s∈S |p − s|S .

Utilization of equilibrium pair of sets.

(Staying therein might have to be paid.)

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 32
Summary of Uncertainty Effect

State and control sets dynamics.

Global Xk+1 = F(Xk , πk ) and Uk = G(Xk , πk ).
Local XN+k+1 = F(XN+k , πf ) and UN+k = G(XN+k , πf ).
Local/Terminal behaviour.
(Unions of sets of) Terminal sets Xf and Uf = G(Xf , πk ).
Fixed–point set pair X = F(X , πf ) and U = G(X , πf ).
State and control sets constraints.
Initial x0 ∈ X0 .
Global Xk ⊆ X and Uk ⊆ U.
Local XN+k ⊆ Xf and UN+k ⊆ Uf .
Generalized cost functions.
Stage `(x, u) := dist(Q, x, X ) + dist(R, u, U).
Terminal Vf (x) := dist(P, x, X ).

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 33
Robust Model Predictive Control (Part I)

Opening

Model Predictive Control

Robust Model Predictive Control

Dissecting Uncertainty Effects

Tube Model Predictive Control

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 34
Tube Model Predictive Control Paradigm

The state tubes are sequences of sets of possible states.

The control tubes are sequences of sets of possible controls.

State and control tubes play role of state and control sequences.

Tubes are induced from the dynamics, uncertainty and control policy.

Optimal tubes are obtained via tube optimal control.

Tube MPC is repetitive utilization of related tube optimal control.

All robust MPC methods result in tubes.

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 35
Tubes Evolution

Inner and Outer Guaranteed Tubes

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 36
Tube Optimal Control
Given an integer N ∈ N and a state x ∈ X select finite sequences of
control laws ΠN−1 := {π0 (·, ·) , π1 (·, ·) , . . . , πN−1 (·, ·)},
control sets UN−1 := {U0 , U1 , . . . , UN−1 }, and
state sets XN := {X0 , X1 , . . . , XN−1 , XN },
which, for each k ∈ {0, 1, . . . , N − 1}, satisfy
x0 ∈ X0 ,
F(Xk , πk ) ⊆ Xk+1 ,
G(Xk , πk ) ⊆ Uk ,
Xk ⊆ X,
Uk ⊆ U, and
XN ∈ Xf ,
PN−1
and which minimize VN (XN , UN−1 ) := k=0 `(Xk , Uk )+ Vf (XN ) .

Parameterization of tubes and control policy is of major importance!

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 37
Robust MPC Personal References

Encyclopedia of Systems and Control

DOI 10.1007/978-1-4471-5102-9_2-1
© Springer-Verlag London 2014

Robust Model-Predictive Control

Saša Raković
Member of the Senior Common Room at St. Edmund Hall, Oxford University, Oxford, UK
Invention of Prediction Structures and
Categorization of Robust MPC Syntheses ⋆
Saša V. Raković ∗
∗
St Edmund Hall, Oxford University, UK
Abstract
Abstract: This plenary paper is concerned with robust model predictive control (MPC)
Model-predictive control (MPC) is indisputably one of the rare modern control techniques that has synthesis. In particular, the novel notion of prediction structures is introduced, and then utilized
to derive a precise and compact overview of the existing robust MPC (RMPC) syntheses as
significantly affected control engineering practice due to its unique ability to systematically handle well as to indicate direct improvements of their system theoretic properties. The prediction
constraints and optimize performance. Robust MPC (RMPC) is an improved form of the nominal structures paradigm allows for a systematic, implementation and examples independent,
MPC that is intrinsically robust in the face of uncertainty. The main objective of RMPC is to devise comparison and classification of the existing RMPC syntheses. The corresponding categorization
an optimization-based control synthesis method that accounts for the intricate interactions of the of the currently available RMPC syntheses is derived by: (i) introducing the adequate
indices as measures of structural (including topological and system theoretic) properties and
uncertainty with the system, constraints, and performance criteria in a theoretically rigorous and computational complexity, and (ii) analysing the trade–off between the guaranteed structural
computationally tractable way. RMPC has become an area of theoretical relevance and practical properties and the necessary computational complexity. The associated analysis is based on
importance but still offers the fundamental challenge of reaching a meaningful compromise the classical game and utility theory notions; for simplicity, it is delivered by deploying
between the quality of structural properties and the computational complexity. the aggregated structural property and computational complexity indices; furthermore, it is
complemented with an unambiguous analysis of the ”holy–grail” trade–off between the quality
of structural properties and the degree of computational complexity.
Keywords Model-predictive control • Robust model-predictive control • Robust optimal control
• Robust stability Keywords: Robust Model Predictive Control, Prediction Structures, Parameterized Tubes.

1. INTRODUCTION The existing approaches to RMPC can be divided broadly

Introduction into two categories depending on the treatment of the un-
Robust model predictive control is an area of theoretical certainty and its interactions with evolution of the system,
relevance and practical importance (Rawlings and Mayne, constraints and performance criteria. The first category
RMPC is an optimization-based approach to the synthesis of robust control laws for constrained of alternatives includes the methods that aim to employ
2009). The field has attracted significant research attention
control systems subject to bounded uncertainty. RMPC synthesis can be seen as an adequately over the last few decades. Nevertheless, the area still offers the inherent robustness, when plausible, of determinis-
defined repetitive decision-making process, in which the underlying decision-making process the fundamental challenge of reaching a reasonable com- tic MPC. These methods employ essentially determinis-
promise between computational practicability and qual- tic MPC synthesis, albeit applied to a suitably modified
is a suitably formulated robust optimal control (ROC) problem. The underlying ROC problem description of the system, constraints and performance
ity of both topological and system theoretic properties.
is specified in such a way so as to ensure that all possible predictions of the controlled state From a theoretical point of view, RMPC synthesis is an criteria, and utilize the notion of input–to–state stability.
and corresponding control actions sequences satisfy constraints and that the “worst-case” cost adequately defined repetitive decision making process, in This category of the alternative approaches to RMPC re-
is minimized. The decision variable in the corresponding ROC problem is a control policy (i.e., a which the underlying decision making process is a suitably sults potentially in computationally practicable methods.
formulated robust optimal control (ROC) problem. Thus, However, the presence of uncertainty is taken care of in an
sequence of control laws) ensuring that different control actions are allowed at different predicted indirect way, and furthermore such approaches seem to be
theoretically exact RMPC synthesis can be attained by
states, while the uncertainty takes on a role of the adversary. RMPC utilizes recursively the solution employing, in a repetitive fashion, the dynamic program- limited since the deterministic MPC is itself an inherently
to the associated ROC problem in order to implement the feedback control law that is, in fact, equal ming (Bellman, 1957) solution of the underlying ROC fragile (non–robust) process. In this sense, it is known
that the stability property of deterministic MPC is non–
to the first control law of an optimal control policy. problem (Mayne et al., 2000; Bertsekas, 2007). Unfor-
robust (Grimm et al., 2004), but the situation is, in fact,
tunately, the associated computational complexity is, in
A theoretically rigorous approach to RMPC synthesis can be obtained either by employing, in general, impracticable. It is, hence, rather tempting to aim even worse since the optimal control of constrained discrete
a repetitive fashion, the dynamic programming solution of the corresponding ROC problem or by to develop the methods which yield directly approximate time systems is a fragile process itself (Raković, 2009). A
solving online, in a recursive manner, an infinite-dimensional optimization problem (Rawlings and solutions (Jones et al., 2007) to the underlying exact ROC more detailed overview of this category of RMPC synthesis
problem. However, this approach is unwieldy, since the in- methods can be found in (Raimondo et al., 2009). The sec-
Mayne 2009). In either case, the associated computational complexity renders the exact RMPC ond category of alternative approaches to RMPC includes
herited complexity of the traditionally employed minimax
synthesis hardly ever tractable. This computational impracticability of the theoretically exact optimization (Scokaert and Mayne, 1998; Kerrigan and the methods that account for the effects of the uncertainty
RMPC, in conjunction with the convoluted interactions of the uncertainty with the evolution of the Maciejowski, 2003) is bound to induce a reasonably high directly, but aim to exploit suitable parameterization of
degree of computational complexity to any approximation the underlying control policy in order to reduce the cor-
based methodology that would guarantee reasonable sys- responding computational complexity (Blanchini, 1990;
� tem theoretic properties. All in all, the prohibitive compu- Chisci et al., 2001; Löfberg, 2003b,a; Langson et al., 2004;
E-mail: svr@sasavRaković.com
tational complexity of the exact RMPC synthesis (or its Mayne et al., 2005; S. V. Raković, 2005; Goulart et al.,
direct minimax optimization based approximations) has 2006; Raković et al., 2012d). This category of approaches
Page 1 of 11 motivated the development of alternative approaches. has been developed compatibly with a commonly accepted
consensus: there is a need for simplifying approximations
of the underlying control policy and for cost functions that
⋆ Email: [email protected].

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 38
Robust Model Predictive Control (Part I)

To Be Continued ...

Saša V. Raković, Ph.D. DIC Robust Model Predictive Control (Part I) 1st AMPCSS @ UW–Madison 39