Professional Documents
Culture Documents
Lou Jiang 2016 Event Triggered Control of Spatially Distributed Processes Via Unmanned Aerial Vehicle
Lou Jiang 2016 Event Triggered Control of Spatially Distributed Processes Via Unmanned Aerial Vehicle
aerial vehicle
Abstract
This article concerns the problem of event-triggered observer-based output feedback control of spatially distributed
processes under the autonomous operation of an unmanned aerial vehicle. The specific spatially distributed process is
modeled within the distributed parameter systems framework. To control the considered distributed parameter system
efficiently, we first estimate the states with an observer based on the measurement information from sensors. Then, an
event-triggered observer-based controller is designed, which can reduce the frequency of signal transmissions between
the observer and the controller. In contrast to normal sampled-data controller that is updated periodically, the event-
triggered controller is updated only when an ‘‘event’’ happens. Moreover, the Zeno behavior is also excluded by proving
there exists a lower bound for interexecution time. Numerical simulations are finally presented to illustrate the effec-
tiveness of the proposed control method.
Keywords
Unmanned aerial vehicle, event-triggered control, distributed parameter system, observer
Topic: Special Issue - Intelligent Flight Control for Unmanned Aerial Vehicles
Topic Editor: Mou Chen
Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 3.0 License
(http://www.creativecommons.org/licenses/by/3.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 International Journal of Advanced Robotic Systems
(EC) has become hot research topics and drawn consider- wðt; 0Þ ¼ wðt; lÞ ¼ 0 and the initial condition is
able attention,13–16 because it may allow to significantly wð0; xÞ ¼ w 0 ðxÞ 2 L2 ðÞ. The functions bðx; xa ðtÞÞ and
reduce the usage of the communication channel. For uðtÞ denote the spatial distribution and the control input
instance, distributed event-based control approaches have of the UAV, respectively, where xa ðtÞ denotes the loca-
been successfully applied to analyze and design coopera- tion of the mobile UAV.
tive controllers in multiagent systems.17–19 The spatial distribution bðx; xa ðtÞÞ is assumed to be
Compared to time-triggered periodic control, the EC
1 if x 2 ½xa ðtÞ "; xa ðtÞ þ "
possesses the prominent advantage of reducing communi- bðx; xa ðtÞÞ ¼ (2)
cations.13 The important technology of sensor/actuator net- 0 otherwise
works, consisting of flight vehicles 20,21 or mobile where xa ðtÞ and " > 0 are the position and the spatial sup-
agents,18,19 provides radically new communication and net- port of the mobile UAV, respectively.
working paradigms with many new application.22 Never- The measurement output on the DPS (1) is given by
theless, most of the achievements focus on finite
2 ðl 3
dimensional dynamical systems, the mathematical models 2 3 s
of which are expressed by the ordinary differential equa- y 1 ðtÞ 6 cðx; x 1 ðtÞÞwðt; xÞ dx 7
6 0 7
tions, while research on the problem of EC for DPSs with 6 7 6 7
6 .. 7 6 . 7
guidance of unmanned aerial vehicle (UAV) or mobile yðtÞ ¼ 6 . 7 ¼ 6 .. 7 (3)
4 5 6 7
actuators are rare. 6 ðl 7
ym ðtÞ 4 5
The primary goal of this article is to discuss the esti- cðx; xsm ðtÞÞwðt; xÞ dx
mation and control problem in spatially distributed pro- 0
cesses under the autonomous operation of a UAV and where cðx; xsi ðtÞÞ, i¼ 1; 2; . . . ; m denotes the sensor spatial
handle the problem through an event-triggered observer- distribution defined by
based output feedback control strategy providing addi-
tional flexibility and improvement in communication. s 1 if x 2 ½xsi ðtÞ ; xsi ðtÞ þ
cðx; xi ðtÞÞ ¼ (4)
An observer through the UAV is designed to estimate the 0 otherwise
states of the DPSs. Based on the state estimates, an event-
trigger condition is provided to determine when the con- where xsi ðtÞ denotes the position of the i th sensor and > 0
trol information can be transferred to the actuator. To is the spatial support of the i th sensor.
analyze whether the estimation error is asymptotically It is convenient to bring the aforementioned systems (1)
convergent or not, a proper Lyapunov function is con- and (3) in an abstract framework. Let X ¼ L2 ðÞ be a
structed. Finally, the proposed control method is applied Hilbert space with the inner product h; i and correspond-
to a spatially distributed process described by one- ing induced norm j j. The reflexive Banach space V with
dimensional partial differential equation through com- norm denoted by k k is identified by the Sobolev
puter simulations. space V ¼ H01 ¼ f 2 H 1 ðÞj ð0Þ ¼ ðlÞ ¼ 0g, where
This article is organized as follows. In the second sec- V ¼ H 1 ðÞ denotes the conjugate dual of V with induced
tion, we give the mathematical framework for DPSs. The norm k k . It follows that V°X °V with both embedding
third section serves to derive the designed observer and dense and continuously, and as a consequence, we have
event-triggered controller. The fourth section gives the glo- jj p k k; 2 V, for some positive constant p.23,24
bal uniform ultimate boundedness and minimum interevent Using the representation, the parabolic operator A asso-
time (MIEI). The fifth section addresses numerical simula- ciated with (1) is given by24
tions where the proposed control method is compared to the d 2
normal sampled-data control (SC), and finally, conclusion A ¼ a1 a2 ; 2 DomðAÞ (5)
dx 2
remarks are given in the last section.
where
0
Mathematical formulation DomðAÞ ¼ f 2 L2 ðÞ j ; absolutely continuous;
Consider a diffusion process modeled by a parabolic DPS, and ð0Þ ¼ ðlÞ ¼ 0g:
given by
Following the study by Demetriou10 and Demetriou and
@wðt; xÞ @ 2w Hussein,23 the operator A satisfies
¼ a 1 2 ðt; xÞ a 2 wðt; xÞ þ bðx; xa ðtÞÞuðtÞ (1)
@t @x jhA; ’ij k kk ’ k
where a 1 ; a2 > 0, t 2 ½0; 1Þ and x 2 denote the time
and
variable and the spatial variable, respectively.
¼ ½0; l⊂R represents the bounded interval and hA; i a 1 k k2
wðt; xÞ 2 R is the state of the system. It is assumed
that the Dirichlet boundary value conditions are for ; ’ 2 V, where ¼ a1 þ a2 p2 .
Lou and Jiang 3
State estimation process and event- Based on the observer states, the EC input is designed as
triggered controller
^ k Þ;
uðtÞ ¼ Kwðt t 2 ½tk ; tkþ1 Þ (9)
The estimation of spatially distributed processes is
achieved by a UAV gathering information from m sensors. for k 2 f1; 2; . . . 1g; where K is the controller gain oper-
We design a Luenberger observer as follows ^ k Þ is the estimated state or the sampled observer
ator and wðt
state.
^_ ðtÞ ¼ AwðtÞ
w ^ þ Bðxa ðtÞÞuðtÞ þ LðyðtÞ y^ðtÞÞ;
(7)
^
wð0Þ 6¼ w 0 ; Assumption 2. There exists a bounded operator K such that
2 1 > 0, where is the minimum eigenvalue of
where y^ðtÞ ¼ CwðtÞ
^ and L is the observer gain that is taken
A þ Bðxa ðtÞÞK.
as L ¼ rC in this article, where C is the adjoint of the
output operator C and r is a positive constant. Assumption 3. Suppose that jBðxa ðtÞÞKj 2 > 0, where
Assumption 1. There exists a bounded operator L such that is the minimum eigenvalue of A þ rC C, where r is
A LC generates a C0 -semigroup. defined in (7).
For simplicity, the operator K ¼ B ðxa ðtÞÞ is consid-
Remark 1. With the help of the observability condition ered in this article, where > 0 is a constant and B ðxa ðtÞÞ
in assumption 1, the well posedness of (7) can be is the adjoint of the operator Bðxa ðtÞÞ.
established using the fact that A LC generates a This work aims to determine an MIEI T min and improve
C0 -semigroup. the estimation and control performance with guaranteed
Next, we introduce an event-triggered observer-based stability property.
control scheme whose structure is depicted in Figure 1.
The event-triggered threshold is monitored by event
detector that determines when to transmit the newest Main results
state estimates from the observer to the controller. Spe-
cifically, let ftk g1
k¼0 be the time instants when an event
Define the estimation error eðtÞ ¼ wðtÞ wðtÞ.
^ Then, the
happens with tk < tkþ1 . The next instant tkþ1 is deter- error system is given by
mined by _ ¼ ðA rC CÞeðtÞ
eðtÞ
(10)
eð0Þ ¼ wð0Þ wð0Þ
^ 6¼ 0
tkþ1 ¼ inf ft > tk j jwðtÞ
^ wðt
^ k Þj eg (8)
within the time period t 2 ½tk ; tkþ1 Þ.
^ k Þ denotes the esti-
for all k 2 f1; 2; . . . g; where wðt
mated states sampled at time tk and e is the event Theorem 1. Consider the DPS (1) and its observer (7) with
threshold. sampling instants determined by (8). Suppose that
4 International Journal of Advanced Robotic Systems
assumptions 1 and 3 hold, then, the estimation error eðtÞ of Correspondingly, the state observer (7) can be derived as
the error system (10) is globally convergent, that is
^_ ¼ ðA Bðxa ðtÞÞKÞwðtÞ
wðtÞ eðtÞ þ rC CeðtÞ
^ þ Bðxa ðtÞÞK^
lim jeðtÞj ¼ 0 (11) (17)
t!1
^ xÞ ¼ w
with the initial condition wð0; ^ 0 ðxÞ 2 L2 ðÞ for
Proof. Consider a Lyapunov function t 2 ½tk ; tkþ1 Þ.
The following theorem shows the global uniform ulti-
V ðtÞ ¼ heðtÞ; ðA þ rC CÞeðtÞi mate boundedness of the states of the system (1).
Then, the time derivative of V ðtÞ for t 2 ½tk ; tkþ1 Þ along
with (10) gives Theorem 2. Consider the DPS (1) with the event-triggered
controller (9) and its observer (7) with sampling instants
V_ ðtÞ ¼ 2heðtÞ;
_ ðA þ rC CÞeðtÞi determined by (8). Suppose that assumptions 1–3 hold, then
¼ 2jðA rC CÞeðtÞj 2 2jeðtÞj 2 (12) the states of the system (1) are globally uniformly ulti-
mately bounded, that is
where the condition in assumption 3 is used. Then, it fol- sffiffiffiffi
lows that
lim jwðtÞj (18)
lim jeðtÞj ¼ 0 (13)
t!1
t!1
where ¼ minf2 1; 2 2jBðxa ðtÞÞKj 2 g and
This completes the proof. c ¼ 2jBðx ðtÞÞKj 2 e2 .
a
With the control input (9), the actuator in UAV is
utilized to control the DPS. In this paper, the actuator Proof. Consider a Lyapunov function
and controller are assumed to be collocated and the
actuator motion is also event-triggered. When the con- V ðtÞ ¼ hwðtÞ; wðtÞi þ heðtÞ; eðtÞi
troller receives the observer states at time instance tk , it The time derivative of V ðtÞ for t 2 ½tk ; tkþ1 Þ gives
can compute the position where the UAV should move
to. Let xa ðtÞ be the actuator position, then V_ ðtÞ ¼ 2hwðtÞ;
_ wðtÞi þ 2heðtÞ;
_ eðtÞi
xa ðtÞ ¼ xa ðtk Þ; t 2 ½tk ; tkþ1 Þ. At the time instances tk , the ¼ 2hðA Bðxa ðtÞÞKÞwðtÞ þ Bðxa ðtÞÞKeðtÞ
desired actuator position is
eðtÞ; wðtÞi þ 2hðA rC CÞeðtÞ; eðtÞi
þ Bðxa ðtÞÞK^
xa ðtk Þ ¼ arg max jwðt
^ k ; xÞj (14) ¼ 2hðA Bðxa ðtÞÞKÞwðtÞ; wðtÞi þ 2hBðxa ðtÞÞKeðtÞ
x2
eðtÞ; wðtÞi þ 2hðA rC CÞeðtÞ; eðtÞi
þ Bðxa ðtÞÞK^
which gives the location over the domain with the max-
imum value of the observer states. (19)
Using the inequality 2hx; yi h x; xi þ hy; yi, we further
Remark 2. If the actuator is only allowed to move from its obtain
current position xa ðtk Þ with a maximum distance +h,
which means that the actuator moves only in V_ ðtÞ 2hðA Bðxa ðtÞÞKÞwðtÞ; wðtÞi
½xa ðtk Þ h; xa ðtk Þ þ h, the proposed actuator position is þ 2hðA rC CÞeðtÞ; eðtÞi þ jBðxa ðtÞÞKeðtÞ
modified by
eðtÞj2 þ jwðtÞj2
þ Bðxa ðtÞÞK^
a
x ðtk Þ ¼ arg max jwðt
^ k ; xÞj (15) ⭐2hðA Bðxa ðtÞÞKÞwðtÞ; wðtÞi
xa ðtk1 Þhxxa ðtk1 Þþh
þ 2hðA rC CÞeðtÞ; eðtÞi
which finds the location of the maximum value of the
observer states over the domain ½xa ðtk1 Þ h; þ 2jBðxa ðtÞÞKj 2 jeðtÞj2
xa ðtk1 Þ þ h and the actuator will move to the location. eðtÞj2 þ jwðtÞj2
þ 2jBðxa ðtÞÞKj 2 j^ (20)
Next, the stability of the controlled DPS (1) will be
analyzed. The closed-loop system under controller (9) is By assumptions 2 and 3, the positive definite operators
written as A þ Bðxa ðtÞÞK and A þ rC C satisfy
V_ ðtÞ ð1 2ÞjwðtÞj2
þ 2ðjBðxa ðtÞÞKj 2 ÞjeðtÞj2
þ 2jBðxa ðtÞÞKj 2 j^
eðtÞj 2
ðjwðtÞj 2 þ jeðtÞj 2 Þ þ 2jBðxa ðtÞÞKj 2 j^
eðtÞj2
where ¼ minf2 1; 2 2jBðxa ðtÞÞKj 2 g.
Due to the event-triggered condition (8), j^eðtÞj2 e2
holds for all t 2 ½tk ; tkþ1 Þ, so we have
V_ ðtÞ
V ðtÞ þ
¼ 2jBðxa ðtÞÞKj 2 e2 . Applying the comparison
where
25
lemma, it follows that
ðt
V ðtÞ e V ð0Þ þ eðtsÞ ds
t
0
V ð0Þ et þ
t
¼e (23)
which implies that jwðtÞj2 V ðtÞ e t V ð0Þ
Þe
ð=
t Þ. Therefore, we have
þ ð=
sffiffiffiffi
lim jwðtÞj (24)
t!1
The proof is completed. c
To exclude Zeno behavior, interevent times should be
lower bounded away from 0, which is ensured in the fol-
lowing theorem.
is lower bounded. Figure 2. Space–time plots of the system state wðt; xÞ under two
schemes.
Proof. From the definition e^ðtÞ ¼ wðtÞ ^ wðt
^ k Þ, for any
k 2 f1; 2; . . . g and t 2 ½tk ; tkþ1 , we have
12
e^_ ðtÞ ¼ wðtÞ
^_ ^ k Þ þ rC CeðtÞ
^ Bðxa ðtÞÞKwðt
¼ AwðtÞ SC scheme
EC scheme
¼ AwðtÞ
^ Awðt ^ k Þ Bðxa ðtÞÞKwðt
^ k Þ þ Awðt ^ kÞ 10
(a) t = 4 (b) t = 8
0.35 0.14
0.3 0.12
0.25 0.1
w (t, x)
w (t, x)
0.2 0.08
0.15 0.06
0.1 0.04
0.05 0.02
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x x
(c) t = 12 (d) t = 16
0.07 0.035
0.06 0.03
0.05 0.025
w (t, x)
w (t, x)
0.04 0.02
0.03 0.015
0.02 0.01
0.01 0.005
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x x
Figure 4. Spatial distributions of the system state wðt; xÞ at different time instances (EC scheme: blue line, SC scheme: red dashed line).
EC: event-triggered control; SC: sampled-data control.
ðt ðT
j^
eðtÞj ¼ j eAðtsÞ rC CeðsÞ þ ðA Bðxa ðtÞÞKÞwðt
^ k Þ ds Uðtk Þ jeA j d ¼ e (31)
tk 0
ðt
jeAðtsÞ j jrC Cj jeðsÞj þ jðA Bðxa ðtÞÞKÞwðt
^ k Þj ds which implies that the MIEI T min (T min T) is lower
tk
ðt bounded. c
jeAðtsÞ j jrC Cj e þ jðA Bðxa ðtÞÞKÞwðt
^ k Þj ds
tk Remark 3. The results presented in theorems 1–3 are pro-
ðt vided for a diffusion process modeled by the parabolic DPS
¼ Uðtk Þ jeAðtsÞ j ds (1), but extensions to second-order DPSs are possible.11 In
tk
this work, the event detector is simply implemented by an
(28)
event-triggered threshold, but the proposed results can be
where Uðtk Þ ¼ jðA Bðxa ðtÞÞKÞwðt ^ k Þj þ jrC Cj~
e. extended to the situation where different event detectors,
Denote T ¼ t tk . Then, we have such as event-driven detector with exponentially decreas-
ðt ð tk þT ðT ing threshold15 and sampled-data event detector,16 are
jeAðtsÞ
j ds ¼ je Aðtk þTsÞ
j ds ¼ jeA j d (29) imposed.
tk tk 0
0.8
Event instants
0.6
0.4
0.2
0
0 2 4 6 8 10 12 14 16 18 20
t (s)
22. Nayak A and Stojmenovic I. Wireless sensor and actuator 24. Dautray R and Lions JL. Mathematical analysis and numer-
networks: algorithms and protocols for scalable coordination ical methods for science and technology. Vol. 2: functional
and data communication. Hoboken: John Wiley & Sons, 2010. and variational methods, Berlin, Heidelberg and New York:
23. Demetriou MA and Hussein II. Estimation of spatially dis- Springer-Verlag, 2000.
tributed processes using mobile spatially distributed sensor 25. Khalil H. Nonlinear systems. 3rd ed. Chapter 3.4,
network. SIAM J Control Optim 2009; 48(1): 266–291. pp. 102–103. Prentice Hall: Pearson Education, Inc., 2002.