Event-Triggered Tracking Control: a Discrete-Time

Approach

D. Sbarbaro 1 , J. M. Gomes da Silva Jr.2 , L. G. Moreira3

1 Department of Electrical Engineering

Universidad de Concepcion, Concepcion, Chile.

(e-mail: [email protected])
2 Department of Automation and Energy Systems (DELAE)

Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil

(e-mail:[email protected])
3 DEPEX, IFSUL - RS, Brazil.

(e-mail: [email protected])

IFAC World Congress, July 2020

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 1 / 21
 Overview

1 Introduction

2 Discrete-Time Design

3 Event-Triggered Control Problem

Stability Conditions
Tracking Accuracy and Control Events

4 Optimization Problems

5 Numerical Example

6 Concluding Remarks

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 2 / 21
 Introduction
Event-triggered strategies (ETC), are very effective to deal with
bandwidth limitations in shared networks and energy consumption in
embedded and wireless systems.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 3 / 21
 Introduction
Event-triggered strategies (ETC), are very effective to deal with
bandwidth limitations in shared networks and energy consumption in
embedded and wireless systems.
A large number of ETC strategies have been developed in a
continuous-time framework. After design a discretization will be
required, leading to an approximated implementation of the
continuous ETC.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 3 / 21
 Introduction
Event-triggered strategies (ETC), are very effective to deal with
bandwidth limitations in shared networks and energy consumption in
embedded and wireless systems.
A large number of ETC strategies have been developed in a
continuous-time framework. After design a discretization will be
required, leading to an approximated implementation of the
continuous ETC.
Observer-based output-feedback control has been proposed by several
authors for discrete time systems.
Conditions for the stability of the closed-loop system in terms of linear
matrix inequalities (LMIs). Groff et. al. 2016
Convex optimization problems to tune the trigger functions aiming at a
reduction of the number of events.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 3 / 21
 Introduction
Event-triggered strategies (ETC), are very effective to deal with
bandwidth limitations in shared networks and energy consumption in
embedded and wireless systems.
A large number of ETC strategies have been developed in a
continuous-time framework. After design a discretization will be
required, leading to an approximated implementation of the
continuous ETC.
Observer-based output-feedback control has been proposed by several
authors for discrete time systems.
Conditions for the stability of the closed-loop system in terms of linear
matrix inequalities (LMIs). Groff et. al. 2016
Convex optimization problems to tune the trigger functions aiming at a
reduction of the number of events.
Not many references on tracking and disturbance rejection problems.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 3 / 21
 Introduction
Event-triggered strategies (ETC), are very effective to deal with
bandwidth limitations in shared networks and energy consumption in
embedded and wireless systems.
A large number of ETC strategies have been developed in a
continuous-time framework. After design a discretization will be
required, leading to an approximated implementation of the
continuous ETC.
Observer-based output-feedback control has been proposed by several
authors for discrete time systems.
Conditions for the stability of the closed-loop system in terms of linear
matrix inequalities (LMIs). Groff et. al. 2016
Convex optimization problems to tune the trigger functions aiming at a
reduction of the number of events.
Not many references on tracking and disturbance rejection problems.
In this work, an observer-based ETC strategy for tracking and
disturbance rejection is proposed in a discrete-time framework.
D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira
Event-Triggered
(UC) Tracking Control IFAC 2020 3 / 21
 Discrete-Time Design

Consider the following linear discrete-time system:

xs [k + 1] = Axs [k] + Bu[k] + Bd[k],

(1)
ys [k] = Cxs [k],

where xs ∈ Rn , u ∈ R, ys ∈ R, d ∈ R and r ∈ R are the system states,

input, output, disturbance and reference respectively, k ∈ N.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 4 / 21
 Discrete-Time Design

Consider the following linear discrete-time system:

xs [k + 1] = Axs [k] + Bu[k] + Bd[k],

(1)
ys [k] = Cxs [k],

where xs ∈ Rn , u ∈ R, ys ∈ R, d ∈ R and r ∈ R are the system states,

input, output, disturbance and reference respectively, k ∈ N.
Assumption 1
The pairs (A, B) and (C , A) are supposed controllable and observable,
respectively.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 4 / 21
 Discrete-Time Design

Consider the following linear discrete-time system:

xs [k + 1] = Axs [k] + Bu[k] + Bd[k],

(1)
ys [k] = Cxs [k],

where xs ∈ Rn , u ∈ R, ys ∈ R, d ∈ R and r ∈ R are the system states,

input, output, disturbance and reference respectively, k ∈ N.
Assumption 1
The pairs (A, B) and (C , A) are supposed controllable and observable,
respectively.
Disturbance: Reference:
xd [k + 1] = Ad xd [k], xr [k + 1] = Ar xr [k],
(2) (3)
d[k] = Cd xd [k], r [k] = Cr xr [k],

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 4 / 21
 Discrete-Time Design

For the extended (1) and (2), the following discrete-time observer is
considered:
      
xo [k + 1] A BCd xo [k] B
= + u[k]
xdo [k + 1] 0 Ad xdo [k] 0
 
Ko1
− (ys [k] − yo [k]) (4)
Ko2
yo [k] = Cxo [k]
do [k] = Cd xdo [k]

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 5 / 21
 Discrete-Time Design

For the extended (1) and (2), the following discrete-time observer is
considered:
      
xo [k + 1] A BCd xo [k] B
= + u[k]
xdo [k + 1] 0 Ad xdo [k] 0
 
Ko1
− (ys [k] − yo [k]) (4)
Ko2
yo [k] = Cxo [k]
do [k] = Cd xdo [k]

An observed state feedback control law given by

u[k] = Kc (xo [k] − Kr 1 xr [k]) + Kr 2 xr [k] − Cd xdo [k], (5)

where Kc0 ∈ Rn is the feedback gain designed to stabilize the system and
Kr 1 ∈ Rn×nr and Kr 2 ∈ R1×nr are feedforward gains.
D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira
Event-Triggered
(UC) Tracking Control IFAC 2020 5 / 21
 Discrete-Time Design

We suppose that gains Ko , Kc , Kr 1 and Kr 2 have been designed such that

the following assumptions are satisfied.
Assumption 2
Ko and Kc are such that (Ao + Ko Co ) and (A + BKc ) are Schur stable.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 6 / 21
 Discrete-Time Design

We suppose that gains Ko , Kc , Kr 1 and Kr 2 have been designed such that

the following assumptions are satisfied.
Assumption 2
Ko and Kc are such that (Ao + Ko Co ) and (A + BKc ) are Schur stable.

Assumption 3
Kr 1 and Kr 2 verify the following equations:

Kr 1 Ar = AKr 1 + BKr 2 , (6)

0 = Cr − CKr 1 (7)

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 6 / 21
 Discrete-Time Design

We suppose that gains Ko , Kc , Kr 1 and Kr 2 have been designed such that

the following assumptions are satisfied.
Assumption 2
Ko and Kc are such that (Ao + Ko Co ) and (A + BKc ) are Schur stable.

Assumption 3
Kr 1 and Kr 2 verify the following equations:

Kr 1 Ar = AKr 1 + BKr 2 , (6)

0 = Cr − CKr 1 (7)

Under Assumptions 2 and 3, the perfect reference tracking is

asymptotically achieved, i.e. e[k] → 0 as k → ∞.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 6 / 21
 Event-Triggered Control Problem

In event-triggered control, the update of the control signal is transmitted

to the actuator only at the time instants ni , i = 0, 1, 2..., leading to the
following observer-based state feedback control law ∀k ∈ [ni , ni+1 ):

u[k] = u[ni ] = Kc x o [ni ] + Kr 2 xr [ni ] − do [ni ]. (8)

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 7 / 21
 Event-Triggered Control Problem

In event-triggered control, the update of the control signal is transmitted

to the actuator only at the time instants ni , i = 0, 1, 2..., leading to the
following observer-based state feedback control law ∀k ∈ [ni , ni+1 ):

u[k] = u[ni ] = Kc x o [ni ] + Kr 2 xr [ni ] − do [ni ]. (8)

The control update instants are given by the rule:

ni+1 = min{k > ni : f (δx [k], δd [k], δr [k], x o [k]) > 0} (9)

where

δx [k] = x̄o [ni ] − x̄o [k], δd [k] = do [ni ] − do [k],

δr [k] = xr [ni ] − xr [k],
(10)
i. e., δj [k], with j = x, d, r are a measure of the difference between the
variable used for the last control update and the current observed one.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 7 / 21
 Stability Conditions

The challenge consists in designing the function

f (δx [k], δd [k], δr [k], x o [k])

such that the closed-loop system with the control law (8) is asymptotically
stable.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 8 / 21
 Stability Conditions

The challenge consists in designing the function

f (δx [k], δd [k], δr [k], x o [k])

such that the closed-loop system with the control law (8) is asymptotically
stable.
The closed-loop system behavior is described by the following equations:

x o [k + 1] = (A + BKc )x o [k] + BKc δx [k] − Bδd [k]

(11)
+ BKr 2 δr [k] − Ko1 Co x̃[k]
x̃[k + 1] = (Ao + Ko Co )x̃[k], (12)

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 8 / 21
 Stability Conditions

Theorem 1
Consider that Assumptions 2 and 3 are satisfied. If there exist symmetric
positive definite matrices P1 , P2 , Q̄σ , Qδx , Qδd , Qδr and a matrix Qo of
appropriate dimensions such that

(Ao + Ko Co )0 P1 (Ao + Ko Co ) − P1 + Qo < 0 (13)

Φ − diag (P2 , Qδx , Qo , Qδd , Qδr , Q̄σ ) < 0 (14)

is verified, with
 
ψ11 ψ12 ψ13 ψ14 ψ15 I
 ∗ ψ22 ψ23 ψ24 ψ25 0
 
 ∗ ∗ ψ33 ψ34 ψ35 0
Φ=
 ∗
, (15)
 ∗ ∗ ψ44 ψ45 0
 ∗ ∗ ∗ ∗ ψ55 0
I 0 0 0 0 0
D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira
Event-Triggered
(UC) Tracking Control IFAC 2020 9 / 21
 Stability Conditions

Theorem 1 cont..

ψ11 = (A + BKc )0 P2 (A + BKc ), ψ12 = (A + BKc )0 P2 BKc , ψ13 = −(A + BKc )0 P2 Ko1 Co ,
ψ14 = −(A + BKc )0 P2 B, ψ15 = (A + BKc )0 P2 BKr 2 , ψ22 = (BKc )0 P2 BKc ,
ψ23 = −(BKc )0 P2 Ko1 Co , ψ24 = −(BKc )0 P2 B, ψ25 = (BKc )0 P2 BKr 2 ,
ψ33 = Co0 Ko1
0 P K C ,
2 o1 o ψ34 = Co0 Ko1
0 P B,
2 ψ35 = −Co0 Ko1
0 P BK ,
2 r2
ψ44 = B 0 P2 B, ψ45 = −B 0 P2 BKr 2 , ψ55 = (BKr 2 )0 P2 (BKr 2 )

then the trajectories of the closed-loop system (11)-(12), under the

event-triggered strategy proposed in (9), with the trigger function given by

f (δx [k], δd [k], δr [k], x̄o [k]) = δx [k]0 Qδx δx [k] + δd [k]0 Qδd δd [k]
(16)
+ δr [k]0 Qδr δr [k] − x o [k]0 Qσ x o [k].

where Qσ = Q̄σ−1 , converge asymptotically to the origin, which implies

that the tracking error e[k] → 0 as k → ∞.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 10 / 21
 Stability Conditions

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 11 / 21
 Tracking Accuracy and Control Events
If the disturbance or reference are not constant signals, in steady state the
event-trigger mechanism will degenerate to the periodic one. This can be
clearly seen in the case of sinusoidal signals.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 12 / 21
 Tracking Accuracy and Control Events
If the disturbance or reference are not constant signals, in steady state the
event-trigger mechanism will degenerate to the periodic one. This can be
clearly seen in the case of sinusoidal signals.
It can be practical to consider a trade-off between obtaining less control
updates and tolerating a small tracking error in steady state.
To relax the triggering conditions we consider a real constant γ > 0 so that
fγ = f (δx [k], δd [k], δr [k], x o [k]) − γ > 0

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 12 / 21
 Tracking Accuracy and Control Events
If the disturbance or reference are not constant signals, in steady state the
event-trigger mechanism will degenerate to the periodic one. This can be
clearly seen in the case of sinusoidal signals.
It can be practical to consider a trade-off between obtaining less control
updates and tolerating a small tracking error in steady state.
To relax the triggering conditions we consider a real constant γ > 0 so that
fγ = f (δx [k], δd [k], δr [k], x o [k]) − γ > 0
In this case, we will guarantee that the trajectories are ultimately bounded
in an invariant ellipsoidal set
E(P, η) = {x ∈ R 2n , x 0 Px ≤ η} (17)
with x = [x̃ 0 x̄o0 ]0 and  
P1 0
P= (18)
0 P2
As a consequence, we have that the tracking error, which is given by
−C x̄o [k] will be ultimately bounded as well.
D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira
Event-Triggered
(UC) Tracking Control IFAC 2020 12 / 21
 Tracking Accuracy and Control Events
Theorem 2
Consider that Assumptions 2 and 3 are satisfied. If there exist positive
scalars τ and γ̄, symmetric positive definite matrices P1 , P2 , Q̄σ , Qδx , Qδd ,
Qδr and a matrix Qo of appropriate dimensions such that
(Ao + Ko Co )0 P1 (Ao + Ko Co ) − (1 − τ )P1 + Qo < 0 (19)


Φ − diag (1 − τ )P2 , Qδx , Qo , Qδd , Qδr , Q̄σ < 0 (20)
 
−τ η 1
<0 (21)
1 −γ̄

with Φ defined as in (15), then the trajectories of the closed-loop system

(11)-(12), under the event-triggered strategy proposed in (9) with the
trigger function given by
fγ (δx [k],δd [k], δr [k], x̄o [k]) = δx [k]0 Qδx δx [k] + δd [k]0 Qδd δd [k]
(22)
+ δr [k]0 Qδr δr [k] − x̄o [k]0 Qσ x̄o [k] − γ,
with Qσ = Q̄σ−1 and γ = γ̄ −1 , converge in finite time to the positively
invariant set E(P, η).
D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira
Event-Triggered
(UC) Tracking Control IFAC 2020 13 / 21
 Optimization Problems

A set of suitable matrices Q̄σ , Qδx , Qδd and Qδr aiming at reducing the
number of generated events can be obtained by solving:
Optimization Problem 1

minimize tr(Qδx ) + tr(Qδd ) + tr(Qδr ) + tr(Q̄σ )

(23)
subject to: (13), (14)

with Qδx , Qδr , Qδd and Q̄σ free positive definite matrices of appropriate
dimensions.
This problem is convex, since (13) and (14) are LMIs on the decision
variables.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 14 / 21
 Optimization Problems
For the relaxed triggering condition (22) the parameter γ has to be
appropriately chosen. Let us consider the following constraint
 
0 0
P −η ≥0 (24)
0 σC 0 C

to ensure that E(P, η) is included in the set

Sσ = {x ∈ R2n ; x o C 0 C x o ≤ σ −1
√}, which implies that the tracking error
is confined in a ball with radius σ −1 .
Optimization Problem 2

minimize tr(Qδx ) + tr(Qδd ) + tr(Qδr ) + tr(Q̄σ ) + γ̄

(25)
subject to: (19), (20), (21), (24)

Problem 2 is not convex due to the products τ P and τ η. An optimal

solution can be obtained through the solution of LMI problems by iteration
over the scalar τ .
D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira
Event-Triggered
(UC) Tracking Control IFAC 2020 15 / 21
 Numerical Example
Consider system (1), with the following matrices:
   
1.0000 0.0975 0.0025  
A= , B= , C= 1 0
0 0.9512 0.0488

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 16 / 21
 Numerical Example
Consider system (1), with the following matrices:
   
1.0000 0.0975 0.0025  
A= , B= , C= 1 0
0 0.9512 0.0488

We consider the tracking of a constant reference and the rejection of a

sinusoidal disturbance of frequency ω = 0.25π. In this case it follows that:
 
0.9969 0.0999  
Ar = Cr = 1, Ad = , Cd = 1 0 .
−0.0628 0.9969

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 16 / 21
 Numerical Example
Consider system (1), with the following matrices:
   
1.0000 0.0975 0.0025  
A= , B= , C= 1 0
0 0.9512 0.0488

We consider the tracking of a constant reference and the rejection of a

sinusoidal disturbance of frequency ω = 0.25π. In this case it follows that:
 
0.9969 0.0999  
Ar = Cr = 1, Ad = , Cd = 1 0 .
−0.0628 0.9969

The state feedback and observer gains satisfying Assumption 2 are given
by:  0    
9.2269 1.3449 28.1418
Kc = , Ko1 = , Ko2 = ,
7.7617 5.7430 15.6801
The feedforwardgains Kr 1 and Kr 2 , satisfying the regulator equations (6)
 0
are Kr 1 = 1 0 , Kr 2 = 0.
D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira
Event-Triggered
(UC) Tracking Control IFAC 2020 16 / 21
 Numerical Example

We consider first the conditions given in Theorem 1, which ensures the

perfect tracking and rejection. In this case, to compute the triggering
matrices we solve the Optimization problem 1.
   
1.03 −0.000224 1.3 1.09
Qσ = , Q δx =
−0.000224 0.7 1.09 0.918

Qδd = 0.184, Qδr = 0.000588

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 17 / 21
 Disturbance
0

Numerical Example
−5
0 50 100 150 200 250 300

10
0.5
Tracking error (e)

5
0
0
−0.5
100 150 200 250 300
−5

−10
0 50 100 150 200 250 300

1000

10
500
Control (u)

0 −10
100 150 200 250 300

−500
0 50 100 150 200 250 300

15
Control update events

10

0
0 50 100 150 200 250 300

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 18 / 21
 Numerical Example

Let us now consider the result stated in Theorem 2, with the relaxed
trigger function (22). In this case, for σ = 1, the parameters of the trigger
function obtained from the solution of Optimization problem 2.
   
0.871 −0.0462 1.99 1.67
Qσ = , Q δx =
−0.0462 0.841 1.67 1.41

Qδd = 0.282, Qδr = 0.000604, γ = 0.749

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 19 / 21
 Disturbance
0

Numerical Example
−5
0 50 100 150 200 250 300

10
0.5
Tracking error (e)

5
0
0
−0.5
100 150 200 250 300
−5

−10
0 50 100 150 200 250 300

1000

10
500
Control (u)

0 −10
100 150 200 250 300

−500
0 50 100 150 200 250 300

25
Control update events

20

15

10

0
0 50 100 150 200 250 300

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 20 / 21
 Concluding Remarks

An observer-based event-triggered control control strategy and

stability conditions expressed as a set of LMIs has been proposed to
deal with time varying references and disturbances.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 21 / 21
 Concluding Remarks

An observer-based event-triggered control control strategy and

stability conditions expressed as a set of LMIs has been proposed to
deal with time varying references and disturbances.
To address the trade-off between the reduction of the control updates
and the tolerance to a small steady state error, a relaxed triggering
condition was proposed.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 21 / 21
 Concluding Remarks

An observer-based event-triggered control control strategy and

stability conditions expressed as a set of LMIs has been proposed to
deal with time varying references and disturbances.
To address the trade-off between the reduction of the control updates
and the tolerance to a small steady state error, a relaxed triggering
condition was proposed.
Optimization problems were proposed to compute the parameters of
the triggering functions to reduce the control updates while ensuring
the perfect or the practical tracking.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 21 / 21
 Concluding Remarks

An observer-based event-triggered control control strategy and

stability conditions expressed as a set of LMIs has been proposed to
deal with time varying references and disturbances.
To address the trade-off between the reduction of the control updates
and the tolerance to a small steady state error, a relaxed triggering
condition was proposed.
Optimization problems were proposed to compute the parameters of
the triggering functions to reduce the control updates while ensuring
the perfect or the practical tracking.
Extensions to deal with the multivariable case and parametric
uncertainties.

D. Sbarbaro , J. M. Gomes da Silva Jr., L. G. Moreira

Event-Triggered
(UC) Tracking Control IFAC 2020 21 / 21